fix: 终极状态回滚 + 三个 panel 短路径导包修复

- water_quality_gui_v2.py：回滚机制由 panel_registry.get_tab_index 改为基于 PANEL_REGISTRY[current_tab_idx]['step_id'] 的索引反查，消除不存在的 get_tab_index 导致回滚期间二次崩溃并被静默吞噬的幽灵跳步根因

.gitignore

@@ -155,3 +155,53 @@ tmp/
 *.bak
 *.backup
 *~
+
+# ============================================================
+# 不应进入版本控制的文件类型
+# ============================================================
+
+# Qwen Code 用户配置（个人环境，每次 clone 都不同）
+.qwen/settings.json
+.qwen/settings.json.orig
+
+# Qwen Code 自动生成的 skill 文件（每次会话重新生成）
+.qwen/skills/
+
+# GUI 运行时生成的文件
+src/gui/scaler_params.pkl
+src/gui/crash_dump.txt
+
+# 临时/调试脚本（根目录）
+降采样光谱.py
+1.py
+tset.py
+
+# 报告与文档（本地工作产物）
+封装问题分析报告.md
+软件说明.md
+软件说明2.md
+
+# 数据子目录中非 .gitkeep 的生成文件
+data/sub/waterindex*.csv
+data/sub/waterindex*.xlsx
+data/sub/png/watermask.png
+
+# 图标文件（仅需保留 vector/svg，删除像素图标压缩包副本）
+data/icons-1/
+data/icons/
+
+# 旧版脚手架（遗留实验代码）
+new/
+
+# 精确放行 src/new/（端到端模块化新架构）
+!/src/new/
+!/src/new/**
+!/src/new/core/**
+!/src/new/services/**
+!/src/new/views/**
+
+# 前端脚手架（未集成的独立 Vue 项目）
+frontend/
+
+# 面板备份目录（运行中自动生成）
+_archive_panels_backup_/

1.py

@@ -1,4 +0,0 @@
-
-
-new_wavelengths = [np.mean(wavelengths[i:i+3]) for i in range(0, len(wavelengths), 3)]
-print(new_wavelengths)

README_new_arch.md

@@ -0,0 +1,83 @@
+# 端到端模块化新架构（src/new/）
+
+## 目录结构
+
+```
+src/new/
+├── __init__.py
+├── core/
+│   ├── __init__.py
+│   └── base_view.py          # 基础通讯接口（继承 QWidget + dispatch_execute）
+├── services/                 # 独立后端大脑
+│   ├── __init__.py
+│   ├── step1_service.py      # Step 1 真实服务（execute_step1）
+│   └── placeholder_service.py # step2-step13 占位服务
+├── views/                    # 独立前端皮囊
+│   ├── __init__.py
+│   ├── step1_view.py         # Step 1 真实视图（继承 BaseView）
+│   └── placeholder_view.py   # step2-step13 占位视图
+└── main_view.py              # 路由与调度壳（QMainWindow + QThread）
+```
+
+## 端到端调用链
+
+```
+Step1View._on_run_clicked  (绿色按钮)
+   │  self.dispatch_execute("step1", self.get_config())
+   ▼
+BaseView.dispatch_execute   (沿父链上溯)
+   │  ancestor.run_single_step(step_id, config)
+   ▼
+MainView.run_single_step  (查 ROUTES 表 → 注入 work_dir)
+   │  TaskWorker(service_func, config).start()
+   ▼
+services.step1_service.execute_step1(config)
+   │  调 WaterMaskStep.run(...) → 包装成 dict 返回
+   ▼
+MainView._on_step_done  (按 status 写日志)
+```
+
+## 运行验证
+
+### 1. 三层冒烟（推荐先跑）
+
+```cmd
+cd D:\111\office\ZHLduijie\1.WQ\WQ_GUI
+python _smoke_new_arch.py
+```
+
+预期输出 `汇总：54/54 通过`。
+
+### 2. 启动路由主窗口
+
+```cmd
+cd D:\111\office\ZHLduijie\1.WQ\WQ_GUI
+python -m src.new.main_view
+```
+
+或：
+
+```cmd
+python src\new\main_view.py
+```
+
+启动后：
+
+* 左侧 `QListWidget` 显示 13 个 step（step1 真实，其余占位）
+* 点击 `执行 Step 1: 水域掩膜` → 绿色按钮 → `dispatch_execute`
+* 底部 `QTextEdit` 实时打印 `[Router]` / `[Service]` 日志
+
+## 关键设计原则
+
+1. **view 零业务**：`src/new/views/*.py` 绝不 import 任何 `src/core/`、`src/services/`
+2. **service 零 PyQt**：`src/new/services/*.py` 不 import 任何 PyQt、不读写全局
+3. **唯一跨界通道**：`BaseView.dispatch_execute` 把 (step_id, config) 推给主窗口
+4. **后台执行不阻塞 UI**：`TaskWorker(QThread)` 子线程跑 service
+5. **错误兜底**：service 任何异常都被 TaskWorker 捕获并转成 `{status: "error", ...}`
+
+## 当前状态
+
+| step   | view                | service                | 状态                |
+|--------|---------------------|------------------------|---------------------|
+| step1  | `Step1View` 真实    | `execute_step1` 真实   | ✅ 已迁移           |
+| step2-13 | `PlaceholderView` | `execute_placeholder`  | 🚧 占位待迁移       |

_check_qaa.py

@@ -0,0 +1,4 @@
+import sys
+sys.path.insert(0, r'D:\111\office\ZHLduijie\1.WQ\WQ_GUI')
+from src.core.algorithms.qaa import QAABaselineSolver
+print("QAABaselineSolver imported OK")

check_lines.py

@@ -0,0 +1,6 @@
+import sys
+with open(r'D:\111\office\ZHLduijie\1.WQ\WQ_GUI\src\gui\water_quality_gui.py', 'rb') as f:
+    content = f.read()
+lines = content.split(b'\r\n')
+for i, line in enumerate(lines[2918:2955], start=2919):
+    sys.stdout.buffer.write(f'{i}: {repr(line[:120])}'.encode('utf-8') + b'\n')

data/sub/waterindex.csv

@@ -1,46 +0,0 @@
-Formula_Name,Category,Formula,Reference
-BGA_Am09KBBI,Phycocyanin (BGA_PC),(w686 - w658) / (w686 + w658),"Amin, R.; Zhou, J.; Gilerson, A.; Gross, B.; Moshary, F.; Ahmed, S.; Novel optical techniques for detecting and classifying toxic dinoflagellate Karenia brevis blooms using satellite imagery, Optics Express, 2009, 17, 11, 1-13."
-BGA_Be162B643sub629,Phycocyanin (BGA_PC),w644 - w629,"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 538."
-BGA_Be162B700sub601,Phycocyanin (BGA_PC),w700 - w601,"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 539."
-BGA_Be162BsubPhy,Phycocyanin (BGA_PC),w715 - w615,"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 540."
-BGA_Be16FLHBlueRedNIR,Phycocyanin (BGA_PC),w658 - (w857 + (w458 - w857)),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 538."
-BGA_Be16FLHGreenRedNIR,Phycocyanin (BGA_PC),w658 - (w857 + (w558 - w857)),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 539."
-BGA_Be16FLHVioletRedNIR,Phycocyanin (BGA_PC),w658 - (w857 + (w444 - w857)),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 538."
-BGA_Be16MPI,Phycocyanin (BGA_PC),(w615 - w601) - (w644 - w601),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 539."
-BGA_Be16NDPhyI,Phycocyanin (BGA_PC),(w700 - w622) / (w700 + w622),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 540."
-BGA_Be16NDPhyI644over615,Phycocyanin (BGA_PC),(w644 - w615) / (w644 + w615),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 541."
-BGA_Be16NDPhyI644over629,Phycocyanin (BGA_PC),(w644 - w629) / (w644 + w629),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 542."
-BGA_Be16Phy2BDA644over629,Phycocyanin (BGA_PC),w644 / w629,"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 545."
-BGA_Da052BDA,Phycocyanin (BGA_PC),w714 / w672,"Wynne, T. T., Stumpf, R. P., Tomlinson, M. C., Warner, R. A., Tester, P. A., Dyble, J.; Relating spectral shape to cyanobacterial blooms in the Laurentian Great Lakes. Int. J. Remote Sens., 2008, 29, 3665-3672."
-BGA_Go04MCI,Phycocyanin (BGA_PC),w709 - w681 - (w753 - w681),"Gower, J.F.R.; Brown,L.; Borstad, G.A.; Observation of chlorophyll fluorescence in west coast waters of Canada using the MODIS satellite sensor. Can. J. Remote Sens., 2004, 30 (1), 17<31><37>?5."
-BGA_HU103BDA,Phycocyanin (BGA_PC),(((1 / w615) - (1 / w600)) - w725),"Hunter, P.D.; Tyler, A.N.; Willby, N.J.; Gilvear, D.J.; The spatial dynamics of vertical migration by Microcystis aeruginosa in a eutrophic shallow lake: A case study using high spatial resolution time-series airborne remote sensing. Limn. Oceanogr. 2008, 53, 2391-2406"
-BGA_Ku15PhyCI,Phycocyanin (BGA_PC),(-1 * (W681 - W665 - (W709 - W665))),"Kudela, R.M., Palacios, S.L., Austerberry, D.C., Accorsi, E.K., Guild, L.S.; Application of hyperspectral remote sensing to cyanobacterial blooms in inland waters, Torres-Perez, J., 2015, Remote Sens. Environ., 2015, 167, 1-10."
-BGA_Ku15SLH,Phycocyanin (BGA_PC),(w715 - w658) + (w715 - w658),"Kudela, R.M., Palacios, S.L., Austerberry, D.C., Accorsi, E.K., Guild, L.S.; Application of hyperspectral remote sensing to cyanobacterial blooms in inland waters, Torres-Perez, J., 2015, Remote Sens. Environ., 2015, 167, 1-11."
-BGA_MI092BDA,Phycocyanin (BGA_PC),w700 / w600,"Mishra, S.; Mishra, D.R.; Schluchter, W. M., A novel algorithm for predicting PC concentrations in cyanobacteria: A proximal hyperspectral remote sensing approach. Remote Sens., 2009, 1, 758<35><38>?75."
-BGA_MM092BDA,Phycocyanin (BGA_PC),w724 / w600,"Mishra, S.; Mishra, D.R.; Schluchter, W. M., A novel algorithm for predicting PC concentrations in cyanobacteria: A proximal hyperspectral remote sensing approach. Remote Sens., 2009, 1, 758<35><38>?76."
-BGA_MM12NDCIalt,Phycocyanin (BGA_PC),(w700 - w658) / (w700 + w658),"Mishra, S.; Mishra, D.R.; A novel remote sensing algorithm to quantify phycocyanin in cyanobacterial algal blooms, Env. Res. Lett., 2014, 9 (11), DOI:10.1088/1748-9326/9/11/114003"
-BGA_MM143BDAopt,Phycocyanin (BGA_PC),((1 / w629) - (1 / w659)) * w724,"Mishra, S.; Mishra, D.R.; A novel remote sensing algorithm to quantify phycocyanin in cyanobacterial algal blooms, Env. Res. Lett., 2014, 9 (11), DOI:10.1088/1748-9326/9/11/114004"
-BGA_SI052BDA,Phycocyanin (BGA_PC),w709 / w620,"Simis, S. G. H.; Peters, S.W. M.; Gons, H. J.; Remote sensing of the cyanobacteria pigment phycocyanin in turbid inland water. Limn. Oceanogr., 2005, 50, 237<33><37>?45"
-BGA_SM122BDA,Phycocyanin (BGA_PC),w709 / w600,"Mishra, S. Remote sensing of cyanobacteria in turbid productive waters, PhD Dissertation. Mississippi State University, USA. 2012."
-BGA_SY002BDA,Phycocyanin (BGA_PC),w650 / w625,"Schalles, J.; Yacobi, Y. Remote detection and seasonal patterns of phycocyanin, carotenoid and chlorophyll-a pigments in eutrophic waters. Archiv fur Hydrobiologie, Special Issues Advances in Limnology, 2000, 55,153<35><33>?68"
-BGA_Wy08CI,Phycocyanin (BGA_PC),(-1 * (W686 - W672 - (W715 - W672))),"Wynne, T. T., Stumpf, R. P., Tomlinson, M. C., Warner, R. A., Tester, P. A., Dyble, J.; Relating spectral shape to cyanobacterial blooms in the Laurentian Great Lakes. Int. J. Remote Sens., 2008, 29, 3665-3672."
-Chl_Al10SABI,chlorophyll_a,(w857 - w644) / (w458 + w529),"Alawadi, F. Detection of surface algal blooms using the newly developed algorithm surface algal bloom index (SABI). Proc. SPIE 2010, 7825."
-Chl_Am092Bsub,chlorophyll_a,w681 - w665,"Amin, R.; Zhou, J.; Gilerson, A.; Gross, B.; Moshary, F.; Ahmed, S. Novel optical techniques for detecting and classifying toxic dinoflagellate Karenia brevis blooms using satellite imagery. Opt. Express 2009, 17, 9126<32><36>?144."
-Chl_Be16FLHblue,chlorophyll_a,w529 - (w644 + (w458 - w644)),"Beck, R.A. and 22 others; Comparison of satellite reflectance algorithms for estimating chlorophyll-a in a temperate reservoir using coincident hyperspectral aircraft imagery and dense coincident surface observations, Remote Sens. Environ., 2016, 178, 15-30."
-Chl_Be16FLHviolet,chlorophyll_a,w529 - (w644 + (w429 - w644)),"Beck, R.A. and 22 others; Comparison of satellite reflectance algorithms for estimating chlorophyll-a in a temperate reservoir using coincident hyperspectral aircraft imagery and dense coincident surface observations, Remote Sens. Environ., 2016, 178, 15-30."
-Chl_Be16NDTIblue,chlorophyll_a,(w658 - w458) / (w658 + w458),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 543."
-Chl_Be16NDTIviolet,chlorophyll_a,(w658 - w444) / (w658 + w444),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 544."
-Chl_De933BDA,chlorophyll_a,w600 - w648 - w625,"Dekker, A.; Detection of the optical water quality parameters for eutrophic waters by high resolution remote sensing, Ph.D. thesis, 1993, Free University, Amsterdam."
-Chl_Gi033BDA,chlorophyll_a,((1 / w672) - (1 / w715)) * w757,"Gitelson, A.A.; U. Gritz, and M. N. Merzlyak.; Relationships between leaf chlorophyll content and spectral reflectance and algorithms for non-destructive chlorophyll assessment in higher plant leaves. J. Plant Phys. 2003, 160, 271-282."
-Chl_Kn07KIVU,chlorophyll_a,(w458 - w644) / w529,"Kneubuhler, M.; Frank T.; Kellenberger, T.W; Pasche N.; Schmid M.; Mapping chlorophyll-a in Lake Kivu with remote sensing methods. 2007, Proceedings of the Envisat Symposium 2007, Montreux, Switzerland 23<32><33>?7 April 2007 (ESA SP-636, July 2007)."
-Chl_MM12NDCI,chlorophyll_a,(w715 - w686) / (w715 + w686),"Mishra, S.; and Mishra, D.R. Normalized difference chlorophyll index: A novel model for remote estimation of chlorophyll-a concentration in turbid productive waters, Remote Sens. Environ., 2012, 117, 394-406"
-Chl_Zh10FLH,chlorophyll_a,w686 - (w715 + (w672 - w751)),"Zhao, D.Z.; Xing, X.G.; Liu, Y.G.; Yang, J.H.; Wang, L. The relation of chlorophyll-a concentration with the reflectance peak near 700 nm in algae-dominated waters and sensitivity of fluorescence algorithms for detecting algal bloom. Int. J. Remote Sens. 2010, 31, 39-48"
-Turb_Be16GreenPlusRedBothOverViolet,Turbidity,(w558 + w658) / w444,"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 538"
-Turb_Be16RedOverViolet,Turbidity,w658 / w444,"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 539"
-Turb_Bow06RedOverGreen,Turbidity,w658 / w558,"Bowers, D. G., and C. E. Binding. 2006. 闁炽儲缈籬e Optical Properties of Mineral Suspended Particles: A Review and Synthesis.<2E><>?Estuarine Coastal and Shelf Science 67 (1<><31>?): 219<31><39>?30. doi:10.1016/j.ecss.2005.11.010"
-Turb_Chip09NIROverGreen,Turbidity,w857 / w558,"Chipman, J. W.; Olmanson, L.G.; Gitelson, A.A.; Remote sensing methods for lake management: A guide for resource managers and decision-makers. 2009."
-Turb_Dox02NIRoverRed,Turbidity,w857 / w658,"Doxaran, D., Froidefond, J.-M.; Castaing, P. ; A reflectance band ratio used to estimate suspended matter concentrations in sediment-dominated coastal waters, Remote Sens., 2002, 23, 5079-5085"
-Turb_Frohn09GreenPlusRedBothOverBlue,Turbidity,(w558 + w658) / w458,"Frohn, R. C., & Autrey, B. C. (2009). Water quality assessment in the Ohio River using new indices for turbidity and chlorophyll-a with Landsat-7 Imagery. Draft Internal Report, US Environmental Protection Agency."
-Turb_Harr92NIR,Turbidity,w857,"Schiebe F.R., Harrington J.A., Ritchie J.C. Remote-Sensing of Suspended Sediments闁炽儲鏁刪e Lake Chicot, Arkansas Project. Int. J. Remote Sens. 1992;13:1487<38><37>?509"
-Turb_Lath91RedOverBlue,Turbidity,w658 / w458,"Lathrop, R. G., Jr., T. M. Lillesand, and B. S. Yandell, 1991. Testing the utility of simple multi-date Thematic Mapper calibration algorithms for monitoring turbid inland waters. International Journal of Remote Sensing"
-Turb_Moore80Red,Turbidity,w658,"Moore, G.K., Satellite remote sensing of water turbidity, Hydrological Sciences, 1980, 25, 4, 407-422"

data/sub/waterindex1125.csv

@@ -1,46 +0,0 @@
-Formula_Name,Category,Formula,Reference
-BGA_Am09KBBI,Phycocyanin (BGA_PC),(w686 - w658) / (w686 + w658),"Amin, R.; Zhou, J.; Gilerson, A.; Gross, B.; Moshary, F.; Ahmed, S.; Novel optical techniques for detecting and classifying toxic dinoflagellate Karenia brevis blooms using satellite imagery, Optics Express, 2009, 17, 11, 1-13."
-BGA_Be162B643sub629,Phycocyanin (BGA_PC),w644 - w629,"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 538."
-BGA_Be162B700sub601,Phycocyanin (BGA_PC),w700 - w601,"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 539."
-BGA_Be162BsubPhy,Phycocyanin (BGA_PC),w715 - w615,"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 540."
-BGA_Be16FLHBlueRedNIR,Phycocyanin (BGA_PC),w658 - (w857 + (w458 - w857)),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 538."
-BGA_Be16FLHGreenRedNIR,Phycocyanin (BGA_PC),w658 - (w857 + (w558 - w857)),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 539."
-BGA_Be16FLHVioletRedNIR,Phycocyanin (BGA_PC),w658 - (w857 + (w444 - w857)),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 538."
-BGA_Be16MPI,Phycocyanin (BGA_PC),(w615 - w601) - (w644 - w601),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 539."
-BGA_Be16NDPhyI,Phycocyanin (BGA_PC),(w700 - w622) / (w700 + w622),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 540."
-BGA_Be16NDPhyI644over615,Phycocyanin (BGA_PC),(w644 - w615) / (w644 + w615),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 541."
-BGA_Be16NDPhyI644over629,Phycocyanin (BGA_PC),(w644 - w629) / (w644 + w629),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 542."
-BGA_Be16Phy2BDA644over629,Phycocyanin (BGA_PC),w644 / w629,"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 545."
-BGA_Da052BDA,Phycocyanin (BGA_PC),w714 / w672,"Wynne, T. T., Stumpf, R. P., Tomlinson, M. C., Warner, R. A., Tester, P. A., Dyble, J.; Relating spectral shape to cyanobacterial blooms in the Laurentian Great Lakes. Int. J. Remote Sens., 2008, 29, 3665-3672."
-BGA_Go04MCI,Phycocyanin (BGA_PC),w709 - w681 - (w753 - w681),"Gower, J.F.R.; Brown,L.; Borstad, G.A.; Observation of chlorophyll fluorescence in west coast waters of Canada using the MODIS satellite sensor. Can. J. Remote Sens., 2004, 30 (1), 17<31><37>?5."
-BGA_HU103BDA,Phycocyanin (BGA_PC),(((1 / w615) - (1 / w600)) - w725),"Hunter, P.D.; Tyler, A.N.; Willby, N.J.; Gilvear, D.J.; The spatial dynamics of vertical migration by Microcystis aeruginosa in a eutrophic shallow lake: A case study using high spatial resolution time-series airborne remote sensing. Limn. Oceanogr. 2008, 53, 2391-2406"
-BGA_Ku15PhyCI,Phycocyanin (BGA_PC),-1 * (W681 - W665 - (W709 - W665)),"Kudela, R.M., Palacios, S.L., Austerberry, D.C., Accorsi, E.K., Guild, L.S.; Application of hyperspectral remote sensing to cyanobacterial blooms in inland waters, Torres-Perez, J., 2015, Remote Sens. Environ., 2015, 167, 1-10."
-BGA_Ku15SLH,Phycocyanin (BGA_PC),(w715 - w658) + (w715 - w658),"Kudela, R.M., Palacios, S.L., Austerberry, D.C., Accorsi, E.K., Guild, L.S.; Application of hyperspectral remote sensing to cyanobacterial blooms in inland waters, Torres-Perez, J., 2015, Remote Sens. Environ., 2015, 167, 1-11."
-BGA_MI092BDA,Phycocyanin (BGA_PC),w700 / w600,"Mishra, S.; Mishra, D.R.; Schluchter, W. M., A novel algorithm for predicting PC concentrations in cyanobacteria: A proximal hyperspectral remote sensing approach. Remote Sens., 2009, 1, 758<35><38>?75."
-BGA_MM092BDA,Phycocyanin (BGA_PC),w724 / w600,"Mishra, S.; Mishra, D.R.; Schluchter, W. M., A novel algorithm for predicting PC concentrations in cyanobacteria: A proximal hyperspectral remote sensing approach. Remote Sens., 2009, 1, 758<35><38>?76."
-BGA_MM12NDCIalt,Phycocyanin (BGA_PC),(w700 - w658) / (w700 + w658),"Mishra, S.; Mishra, D.R.; A novel remote sensing algorithm to quantify phycocyanin in cyanobacterial algal blooms, Env. Res. Lett., 2014, 9 (11), DOI:10.1088/1748-9326/9/11/114003"
-BGA_MM143BDAopt,Phycocyanin (BGA_PC),((1 / w629) - (1 / w659)) * w724,"Mishra, S.; Mishra, D.R.; A novel remote sensing algorithm to quantify phycocyanin in cyanobacterial algal blooms, Env. Res. Lett., 2014, 9 (11), DOI:10.1088/1748-9326/9/11/114004"
-BGA_SI052BDA,Phycocyanin (BGA_PC),w709 / w620,"Simis, S. G. H.; Peters, S.W. M.; Gons, H. J.; Remote sensing of the cyanobacteria pigment phycocyanin in turbid inland water. Limn. Oceanogr., 2005, 50, 237<33><37>?45"
-BGA_SM122BDA,Phycocyanin (BGA_PC),w709 / w600,"Mishra, S. Remote sensing of cyanobacteria in turbid productive waters, PhD Dissertation. Mississippi State University, USA. 2012."
-BGA_SY002BDA,Phycocyanin (BGA_PC),w650 / w625,"Schalles, J.; Yacobi, Y. Remote detection and seasonal patterns of phycocyanin, carotenoid and chlorophyll-a pigments in eutrophic waters. Archiv fur Hydrobiologie, Special Issues Advances in Limnology, 2000, 55,153<35><33>?68"
-BGA_Wy08CI,Phycocyanin (BGA_PC),-1 * (W686 - W672 - (W715 - W672)),"Wynne, T. T., Stumpf, R. P., Tomlinson, M. C., Warner, R. A., Tester, P. A., Dyble, J.; Relating spectral shape to cyanobacterial blooms in the Laurentian Great Lakes. Int. J. Remote Sens., 2008, 29, 3665-3672."
-Chl_Al10SABI,chlorophyll_a,(w857 - w644) / (w458 + w529),"Alawadi, F. Detection of surface algal blooms using the newly developed algorithm surface algal bloom index (SABI). Proc. SPIE 2010, 7825."
-Chl_Am092Bsub,chlorophyll_a,w681 - w665,"Amin, R.; Zhou, J.; Gilerson, A.; Gross, B.; Moshary, F.; Ahmed, S. Novel optical techniques for detecting and classifying toxic dinoflagellate Karenia brevis blooms using satellite imagery. Opt. Express 2009, 17, 9126<32><36>?144."
-Chl_Be16FLHblue,chlorophyll_a,w529 - (w644 + (w458 - w644)),"Beck, R.A. and 22 others; Comparison of satellite reflectance algorithms for estimating chlorophyll-a in a temperate reservoir using coincident hyperspectral aircraft imagery and dense coincident surface observations, Remote Sens. Environ., 2016, 178, 15-30."
-Chl_Be16FLHviolet,chlorophyll_a,w529 - (w644 + (w429 - w644)),"Beck, R.A. and 22 others; Comparison of satellite reflectance algorithms for estimating chlorophyll-a in a temperate reservoir using coincident hyperspectral aircraft imagery and dense coincident surface observations, Remote Sens. Environ., 2016, 178, 15-30."
-Chl_Be16NDTIblue,chlorophyll_a,(w658 - w458) / (w658 + w458),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 543."
-Chl_Be16NDTIviolet,chlorophyll_a,(w658 - w444) / (w658 + w444),"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 544."
-Chl_De933BDA,chlorophyll_a,w600 - w648 - w625,"Dekker, A.; Detection of the optical water quality parameters for eutrophic waters by high resolution remote sensing, Ph.D. thesis, 1993, Free University, Amsterdam."
-Chl_Gi033BDA,chlorophyll_a,((1 / w672) - (1 / w715)) * w757,"Gitelson, A.A.; U. Gritz, and M. N. Merzlyak.; Relationships between leaf chlorophyll content and spectral reflectance and algorithms for non-destructive chlorophyll assessment in higher plant leaves. J. Plant Phys. 2003, 160, 271-282."
-Chl_Kn07KIVU,chlorophyll_a,(w458 - w644) / w529,"Kneubuhler, M.; Frank T.; Kellenberger, T.W; Pasche N.; Schmid M.; Mapping chlorophyll-a in Lake Kivu with remote sensing methods. 2007, Proceedings of the Envisat Symposium 2007, Montreux, Switzerland 23<32><33>?7 April 2007 (ESA SP-636, July 2007)."
-Chl_MM12NDCI,chlorophyll_a,(w715 - w686) / (w715 + w686),"Mishra, S.; and Mishra, D.R. Normalized difference chlorophyll index: A novel model for remote estimation of chlorophyll-a concentration in turbid productive waters, Remote Sens. Environ., 2012, 117, 394-406"
-Chl_Zh10FLH,chlorophyll_a,w686 - (w715 + (w672 - w751)),"Zhao, D.Z.; Xing, X.G.; Liu, Y.G.; Yang, J.H.; Wang, L. The relation of chlorophyll-a concentration with the reflectance peak near 700 nm in algae-dominated waters and sensitivity of fluorescence algorithms for detecting algal bloom. Int. J. Remote Sens. 2010, 31, 39-48"
-Turb_Be16GreenPlusRedBothOverViolet,Turbidity,(w558 + w658) / w444,"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 538"
-Turb_Be16RedOverViolet,Turbidity,w658 / w444,"Beck, R.; Xu, M.; Zhan, S.; Liu, H.; Johansen, R.A.; Tong, S.; Yang, B.; Shu, S.; Wu, Q.; Wang, S.; Berling, K.; Murray, A.; Emery, E.; Reif, M.; Harwood, J.; Young, J.; Martin, M.; Stillings, G.; Stumpf, R.; Su, H.; Ye, Z.; Huang, Y. Comparison of Satellite Reflectance Algorithms for Estimating Phycocyanin Values and Cyanobacterial Total Biovolume in a Temperate Reservoir Using Coincident Hyperspectral Aircraft Imagery and Dense Coincident Surface Observations. Remote Sens. 2017, 9, 539"
-Turb_Bow06RedOverGreen,Turbidity,w658 / w558,"Bowers, D. G., and C. E. Binding. 2006. 鈥淭he Optical Properties of Mineral Suspended Particles: A Review and Synthesis.<2E><>?Estuarine Coastal and Shelf Science 67 (1<><31>?): 219<31><39>?30. doi:10.1016/j.ecss.2005.11.010"
-Turb_Chip09NIROverGreen,Turbidity,w857 / w558,"Chipman, J. W.; Olmanson, L.G.; Gitelson, A.A.; Remote sensing methods for lake management: A guide for resource managers and decision-makers. 2009."
-Turb_Dox02NIRoverRed,Turbidity,w857 / w658,"Doxaran, D., Froidefond, J.-M.; Castaing, P. ; A reflectance band ratio used to estimate suspended matter concentrations in sediment-dominated coastal waters, Remote Sens., 2002, 23, 5079-5085"
-Turb_Frohn09GreenPlusRedBothOverBlue,Turbidity,(w558 + w658) / w458,"Frohn, R. C., & Autrey, B. C. (2009). Water quality assessment in the Ohio River using new indices for turbidity and chlorophyll-a with Landsat-7 Imagery. Draft Internal Report, US Environmental Protection Agency."
-Turb_Harr92NIR,Turbidity,w857,"Schiebe F.R., Harrington J.A., Ritchie J.C. Remote-Sensing of Suspended Sediments鈥攖he Lake Chicot, Arkansas Project. Int. J. Remote Sens. 1992;13:1487<38><37>?509"
-Turb_Lath91RedOverBlue,Turbidity,w658 / w458,"Lathrop, R. G., Jr., T. M. Lillesand, and B. S. Yandell, 1991. Testing the utility of simple multi-date Thematic Mapper calibration algorithms for monitoring turbid inland waters. International Journal of Remote Sensing"
-Turb_Moore80Red,Turbidity,w658,"Moore, G.K., Satellite remote sensing of water turbidity, Hydrological Sciences, 1980, 25, 4, 407-422"

data/格式转化.py

@@ -0,0 +1,85 @@
+import os
+from pathlib import Path
+from PIL import Image
+
+
+def batch_convert_to_ico(source_dirs, output_dir, target_size=(256, 256)):
+    """
+    批量将指定目录下的图像文件转换为 ICO 格式。
+
+    :param source_dirs: 包含源文件夹路径的列表
+    :param output_dir: 转换后 ICO 文件的保存目录
+    :param target_size: 输出 ICO 的尺寸，默认 256x256
+    """
+    # 支持的常见输入图像后缀
+    supported_extensions = {'.png', '.jpg', '.jpeg', '.bmp', '.webp', '.tiff'}
+
+    # 确保输出目录存在，若无则自动创建
+    out_path = Path(output_dir)
+    out_path.mkdir(parents=True, exist_ok=True)
+
+    total_converted = 0
+    total_failed = 0
+
+    print("=" * 50)
+    print(f"🚀 开始批量转换 ICO 图标...")
+    print(f"📁 目标输出目录: {out_path}")
+    print("=" * 50)
+
+    # 遍历所有传入的源目录
+    for folder in source_dirs:
+        folder_path = Path(folder)
+
+        if not folder_path.exists():
+            print(f"⚠️ 警告: 源目录不存在，已跳过 -> {folder_path}")
+            continue
+
+        print(f"\n📂 正在扫描目录: {folder_path}")
+
+        # 遍历目录下的所有文件
+        for file_path in folder_path.iterdir():
+            # 仅处理普通文件且后缀在支持列表内（忽略大小写）
+            if file_path.is_file() and file_path.suffix.lower() in supported_extensions:
+                try:
+                    with Image.open(file_path) as img:
+                        # 处理透明通道问题：
+                        # 如果图片支持透明通道 (RGBA/P/LA)，转为 RGBA 确保透明背景不丢失
+                        # 如果是普通 RGB (如 JPG)，转为 RGB
+                        if img.mode in ('RGBA', 'LA') or (img.mode == 'P' and 'transparency' in img.info):
+                            img_clean = img.convert('RGBA')
+                        else:
+                            img_clean = img.convert('RGB')
+
+                        # 构造输出文件名 (原文件名.ico)
+                        new_filename = f"{file_path.stem}.ico"
+                        save_path = out_path / new_filename
+
+                        # 如果目标文件夹中已存在同名文件，为了防止覆盖，可以在文件名后加个标识
+                        # 但通常图标库同名直接覆盖较符合需求，这里默认直接保存
+                        img_clean.save(save_path, format="ICO", sizes=[target_size])
+
+                        print(f"  ✅ 成功: {file_path.name} -> {new_filename}")
+                        total_converted += 1
+
+                except Exception as e:
+                    print(f"  ❌ 失败: 无法转换 {file_path.name}，错误信息: {e}")
+                    total_failed += 1
+
+    print("\n" + "=" * 50)
+    print("🎉 转换任务结束！")
+    print(f"统计: 成功转换 {total_converted} 个文件，失败 {total_failed} 个。")
+    print("=" * 50)
+
+
+if __name__ == "__main__":
+    # 1. 定义你要读取的两个源文件夹路径列表
+    SOURCES = [
+        r"D:\111\office\ZHLduijie\1.WQ\WQ_GUI\data\icons",
+        r"D:\111\office\ZHLduijie\1.WQ\WQ_GUI\data\icons\word"
+    ]
+
+    # 2. 定义统一输出的目标文件夹路径
+    OUTPUT = r"D:\111\office\ZHLduijie\1.WQ\WQ_GUI\data\icons-1"
+
+    # 执行转换
+    batch_convert_to_ico(SOURCES, OUTPUT)

docs/SMOKE_TEST_ROUTE_B_MVP.md

@@ -0,0 +1,350 @@
+# Smoke Test — 路线 B MVP（PipelineContext + AutoML + 软取消 + GUI 缝合）
+
+> 适用范围：路线 B 重构 4 部分（pipeline 包 / AutoML 训练器 / WorkerThread 软取消 / GUI 一键全自动）落盘后的端到端点火试飞清单。
+> 目标：**用最小数据集（1 个 BSQ + 1 个 CSV）在 10–20 分钟内验证全链路打通**。
+
+---
+
+## 0. 前置准备（5 分钟）
+
+### 0.1 装 Optuna
+
+`environment.yml` 当前**未列** optuna（属于本次重构新增依赖）。若不装，Step 6 会自动降级到老 GridSearchCV（仍能跑通，但会触发 fallback 日志）。
+
+```bash
+call venv\Scripts\activate.bat
+pip install "optuna>=3.6,<4.0"
+```
+
+写入 `environment.yml` 的 patch（提交时改）：
+
+```yaml
+  # 路线 B AutoML 防爆引擎（可选；未装时 Step 6 走老 GridSearchCV 降级路径）
+  - optuna>=3.6
+```
+
+### 0.2 准备最小数据集
+
+```text
+work_dir_smoke/
+├── raw/
+│   ├── sample.b              # 假彩色 BSQ（任意小分辨率都行，建议 50×50×6 波段）
+│   ├── sample_mask.tif       # （可选）水域掩膜；不提供则 Step 1 自动生成 NDWI
+│   └── sample.csv            # 含 3–6 个水质参数目标列（Chl-a / TSS / SD / TN / TP / COD…）+ 6 列波段反射率
+└── (其他文件由流程自动生成)
+```
+
+**CSV 模板示例**（`feature_start_column` 默认为第一列；目标列必须**在特征列之前**）：
+
+```csv
+Chl-a,TSS,SD,B1,B2,B3,B4,B5,B6
+12.3,15.1,0.8,0.045,0.052,0.038,0.061,0.072,0.085
+11.8,14.2,0.9,0.044,0.051,0.037,0.060,0.071,0.084
+...  (≥ 200 行；AutoML 智能子采样 N>5000 时才生效)
+```
+
+### 0.3 启动 venv
+
+```bash
+cd /d "D:\111\office\ZHLduijie\1.WQ\WQ_GUI"
+call venv\Scripts\activate.bat
+set PYTHONPATH=src;%PYTHONPATH%
+```
+
+---
+
+## 1. CLI 烟雾（最快路径，3 分钟）— **A 级：必跑**
+
+跳过 GUI，直接验证 `automl_trainer.py` 自身可独立运行 + Optuna 子采样 + 降级路径：
+
+```bash
+python -m src.core.prediction.automl_trainer ^
+    --csv work_dir_smoke/raw/sample.csv ^
+    --feature-start 6 ^
+    --n-trials 5 ^
+    --timeout 60.0 ^
+    --out work_dir_smoke/7_Supervised_Model_Training_AutoML
+```
+
+**通过标准**：
+
+- [ ] 进程退出码 0
+- [ ] 控制台打印 `AutoML: 目标列 X 共尝试 N 个 trial，最佳 CV R²=…`
+- [ ] `<out>/<preprocess>/<target>_<preprocess>_<model>_AUTOML.joblib` 存在
+- [ ] `<out>/automl_summary.json` 存在且 `success=true`
+
+**若 Optuna 未装**，期待看到：
+
+```
+[AutoML] optuna 未安装，全目标列回退老 GridSearchCV
+```
+
+产物文件名带 `_AUTOML` 后缀的逻辑此时**不会触发**（fallback 走老路径），属正常。
+
+---
+
+## 2. GUI 端到端 9 步（核心场景，10–20 分钟）— **S 级：必跑**
+
+### 2.1 启动 GUI
+
+```bash
+call venv\Scripts\activate.bat
+set PYTHONPATH=src;%PYTHONPATH%
+python -m src.gui.water_quality_gui
+```
+
+### 2.2 UI 配置
+
+| 步骤  | 操作                                                                 | 期望                                                                                 |
+| ----- | -------------------------------------------------------------------- | ------------------------------------------------------------------------------------ |
+| 1/9   | 点"选择工作目录" → 选 `work_dir_smoke/`                               | 左侧步骤列表高亮，UI 不报错                                                          |
+| 2/9   | 在 Step 1 面板选 `sample.b`；**掩膜留空**（验证 NDWI 自动生成路径）   | 掩膜文本框保持空白                                                                   |
+| 3/9   | 在 Step 4 面板选 `sample.csv`                                        | CSV 路径显示正确                                                                      |
+| 4/9   | **关键**：其他步骤（2/3/5/5.5/6/7/8/9）保持默认，不改任何参数        | AutoML 默认开启（use_automl=True）                                                   |
+| 5/9   | 点 **▶ 运行完整流程**（不要用老 `run_full_pipeline` 槽）              | 弹出**二次确认窗**，文案显示：<br>• 掩膜：`未指定（将自动生成 NDWI 水域掩膜）`<br>• 去耀斑：开启<br>• AutoML：开启（Optuna 子采样寻优） |
+| 6/9   | 点"是(Y)"                                                            | "运行"按钮变灰，"停止"按钮亮起；进度条归零                                          |
+
+### 2.3 观察日志（重点 4 大检查点）
+
+#### ✅ 检查点 1：ctx 路径传递
+
+启动后**第一秒**应看到类似：
+
+```
+[Runner] ctx 已构造：14 路径字段，4 目录字段
+[Runner] 步骤 1/14：step1_generate_water_mask（requires=['raw_img_path', 'water_mask_path']）
+[Runner] 步骤 2/14：step2_find_glint_area（requires=['raw_img_path', 'water_mask_path', 'output_dir']）
+...
+[Runner] ctx 路径校准：water_mask_path = ...\work_dir_smoke\2_Glint_Area_Mask\glint_mask.tif
+```
+
+→ **若没有 `[Runner]` 日志**，说明 v1 旧路径被走到了，**`inspect.signature` duck-type 没探测到 v2**，回去检查 `worker_thread.py:run()`。
+
+#### ✅ 检查点 2：Step 1 NDWI 自动生成
+
+```
+[Step1] 未指定 mask_path，自动基于 NDWI 生成水域掩膜
+[Step1] NDWI 阈值=0.4，写入 1_Water_Mask/water_mask.tif
+```
+
+→ 验证 `<work_dir>/1_Water_Mask/water_mask.tif` 文件存在且非空。
+
+#### ✅ 检查点 3：AutoML 启用
+
+```
+[Step6] AutoML 启用 Optuna 子采样寻优（timeout=300s, n_trials=20, max_samples=5000）
+[Step6] 目标列 'Chl-a' 共 3 个候选模型，最佳 R²=0.812（model=RandomForest）
+[Step6] 目标列 'TSS'   共 3 个候选模型，最佳 R²=0.745（model=XGBoost）
+[Step6] 训练完成，产物写入 7_Supervised_Model_Training_AutoML/
+[Step6] automl_summary.json 写入完成
+```
+
+→ 验证产物：
+- [ ] `7_Supervised_Model_Training_AutoML/<preprocess>/<target>_<preprocess>_<model>_AUTOML.joblib` ≥ 1 个
+- [ ] `7_Supervised_Model_Training_AutoML/automl_summary.json` 含 `automl: true` 字段
+- [ ] 老目录 `7_Supervised_Model_Training/` **不应该被创建**（AutoML 路径独立）
+
+#### ✅ 检查点 4：AutoML 降级（仅未装 Optuna 时）
+
+```
+[AutoML] optuna 未安装，全目标列回退老 GridSearchCV
+[Step6] 降级路径：调用 WaterQualityModelingBatch.train_models_batch（132 组 GridSearchCV）
+```
+
+→ 跑通即可（仍能产生模型文件），但**降级**属于非优选路径。
+
+### 2.4 9 步全程观察清单
+
+| 步   | 期望产物（路径相对 `work_dir`）                                | 期望耗时（50×50 测试数据） |
+| ---- | -------------------------------------------------------------- | -------------------------- |
+| 1    | `1_Water_Mask/water_mask.tif`                                  | < 5 s                       |
+| 2    | `2_Glint_Area_Mask/glint_mask.tif`                             | < 5 s                       |
+| 3    | `3_Remove_Glint_Image/deglint_image.tif`                       | < 5 s                       |
+| 4    | `4_Process_CSV/processed_data.csv`                             | < 2 s                       |
+| 5    | `5_Training_Sample/training_spectra.csv`                       | < 5 s                       |
+| 5.5  | `5_5_Calculate_Indices/indices.csv`（如启用）                  | < 2 s                       |
+| **6**| `7_Supervised_Model_Training_AutoML/`（**新路径！**）           | **< 5 min（Optuna 5 trial）** |
+| 6.5  | `6_5_Non_Empirical_Modeling/`（如启用）                       | 1–2 min                    |
+| 6.75 | `6_75_Custom_Regression/`（如启用）                            | 1–2 min                    |
+| 7    | `7_Sampling_Points/sampling_points.csv`                        | < 3 s                       |
+| 8    | `8_Prediction/predicted_values.csv`                            | < 5 s                       |
+| 8.5  | `8_5_Prediction_Non_Empirical/predicted.csv`（如启用）          | < 5 s                       |
+| 8.75 | `8_75_Prediction_Custom/predicted.csv`（如启用）               | < 5 s                       |
+| 9    | `9_Kriging_Distribution_Map/distribution_map.tif`              | 5–30 s（纯 Python 慢）      |
+
+### 2.5 流程结束
+
+- [ ] 进度条到 100%
+- [ ] "运行"按钮恢复可点
+- [ ] "停止"按钮变灰
+- [ ] 日志末行出现 `=== 流程执行完成 ===` 或 `=== 流程被取消 ===`（取决于是否点过停止）
+- [ ] 控制台 `on_pipeline_finished` 触发：UI 状态被统一恢复
+
+---
+
+## 3. 软取消测试（3 分钟）— **A 级：必跑**
+
+验证 `threading.Event` 软取消链路（不再用 `terminate()`）。
+
+### 3.1 启动完整流程
+
+如 2.2 启动流程。
+
+### 3.2 中途点"停止"
+
+**时机**：在 Step 6 AutoML 跑 trials 的中途（看到 `[Step6] 目标列 'Chl-a' 共 N 个候选模型` 之后任意时刻）点"停止"。
+
+**期望看到**：
+
+```
+[STOP] 用户请求软取消
+[Step6] 检测到 cancel_event，本 trial 完成后退出
+[Step6] AutoML 在 trial #X 中止，已完成 5/20 trial
+[Runner] 软取消：跳过剩余 8 个 step
+=== 流程被取消 ===
+```
+
+UI 状态：
+
+- [ ] "运行"按钮重新亮起
+- [ ] "停止"按钮变灰
+- [ ] 进度条保留在中断时的百分比（**不**归零）
+- [ ] `on_pipeline_finished` 触发（用 `success=False, cancelled=True` 区分）
+- [ ] **Python 进程不退出**（GUI 仍可继续点"运行"开新流程）
+
+**反例（不应该发生）**：
+
+- ❌ `QThread: Destroyed while thread is still running` 警告
+- ❌ Python 解释器直接崩溃
+- ❌ UI 永远卡死（`run_all_btn` 一直是灰的）
+
+### 3.3 旧 `stop()` 路径回归
+
+为防老代码忘了改，临时把 `water_quality_gui.py:stop_pipeline` 改回 `self.worker.stop()`，跑一次完整流程，看是否出现：
+
+```
+[DEPRECATED] WorkerThread.stop() 已弃用，请改用 soft_stop()。
+```
+
+**这是预期行为**（弃用方法保留但打 warning），流程仍能完成即视为通过。
+
+---
+
+## 4. 失败 / 降级场景（5 分钟）— **B 级：选跑**
+
+### 4.1 未填掩膜 + NDWI 阈值设极端值
+
+把 NDWI 阈值设到 `0.9`（几乎无水域），Step 1 应给出 warning 但不崩：
+
+```
+[Step1] NDWI 阈值=0.9，水域覆盖率 < 1%，请检查影像
+```
+
+### 4.2 CSV 完全无目标列
+
+准备一个**没有目标列的 CSV**（全特征列），点运行：
+
+```
+[AutoML] 训练 CSV 不存在或无目标列：未识别出目标列
+[Step6] AutoML 全部失败，所有目标列返回 success=False
+```
+
+→ UI 不会崩，会在 `automl_summary.json` 写 `error: "未识别出目标列"`。
+
+### 4.3 Step 1 路径不存在
+
+Step 1 选了一个**不存在的 .bsq 文件**：
+
+```
+[Runner] step1_generate_water_mask 异常：FileNotFoundError
+[STOP] 流程中止在 step 1
+```
+
+→ UI 弹错误窗 + 把左侧步骤列表 `setCurrentRow(0)` 自动定位到 Step 1（`_focus_step` 起效）。
+
+### 4.4 Optuna 版本冲突
+
+装一个 `optuna==2.10`（API 大改），跑 GUI：
+
+```
+[AutoML] optuna API 不兼容（>=3.6 要求）：<error>
+[AutoML] 全目标列回退老 GridSearchCV
+```
+
+→ 降级路径生效即视为通过。
+
+---
+
+## 5. 验证矩阵 Checklist
+
+复制以下到 PR 描述 / 验收单：
+
+```markdown
+## 路线 B MVP 验证矩阵
+
+### 代码落盘
+- [ ] src/core/pipeline/__init__.py（17 行，4 export）
+- [ ] src/core/pipeline/context.py（PipelineContext dataclass）
+- [ ] src/core/pipeline/runner.py（StepSpec + PIPELINE_STEPS + PipelineRunner）
+- [ ] src/core/prediction/__init__.py（追加 train_with_automl export）
+- [ ] src/core/prediction/automl_trainer.py（AutoMLResult + train_with_automl + CLI）
+- [ ] src/core/steps/modeling_step.py（use_automl 分支 + _train_models_automl）
+- [ ] src/core/water_quality_inversion_pipeline_GUI.py（run_full_pipeline_v2 + LEGACY_ATTR_MAP + _sync_legacy_attrs_from_context）
+- [ ] src/gui/core/worker_thread.py（cancel_event + soft_stop + run() duck-type）
+- [ ] src/gui/water_quality_gui.py（on_run_all_clicked + _collect_minimal_config + 按钮重连）
+
+### CLI 自测
+- [ ] A.1 `python -m src.core.prediction.automl_trainer --csv ...` 退出码 0
+- [ ] A.2 产物 .joblib 含 `_AUTOML` 后缀
+- [ ] A.3 automl_summary.json 含 success=true
+
+### GUI 端到端
+- [ ] B.1 启动无 ImportError
+- [ ] B.2 二次确认窗文案含 mask 提示 + AutoML 状态
+- [ ] B.3 日志含 [Runner] 前缀（v2 路径生效）
+- [ ] B.4 Step 1 NDWI 自动生成路径生效
+- [ ] B.5 9 步产物路径全部存在
+- [ ] B.6 流程结束后 UI 状态恢复（运行按钮亮、停止按钮灰）
+
+### 软取消
+- [ ] C.1 流程中途点停止，cancel_event 触发
+- [ ] C.2 流程被取消而非崩溃
+- [ ] C.3 UI 状态由 on_pipeline_finished 统一恢复
+- [ ] C.4 旧 stop() 调用打 [DEPRECATED] warning
+
+### 降级
+- [ ] D.1 Optuna 未装 → 全目标列回退老 GridSearchCV
+- [ ] D.2 无目标列 CSV → 写 error 到 summary，不崩 UI
+- [ ] D.3 不存在文件 → _focus_step 定位到对应 step
+```
+
+---
+
+## 6. 已知未做（不在本次范围）
+
+- [ ] Kriging 多进程并行（当前 backend="loop" 纯 Python）
+- [ ] Step 5 radius==0 内存优化（整波段读入）
+- [ ] 进度条 sub-step 粒度（当前只到 step 级）
+- [ ] Step 8 全图预测（当前只对采样点预测）
+- [ ] 全项目搜替换老 `self.worker.stop()` 调用（仅本会话改了 `water_quality_gui.py` 的 stop_pipeline）
+- [ ] `requirements.txt` 同步 Optuna（仅 `environment.yml` 写）
+- [ ] 单元测试套件（`tests/` 目录为空；建议用 pytest 覆盖 train_with_automl / PipelineRunner）
+
+---
+
+## 7. 出问题找哪里
+
+| 现象                                          | 看哪里                                                  |
+| --------------------------------------------- | ------------------------------------------------------- |
+| `[Runner]` 日志没出来                        | `worker_thread.py:run()` 的 `inspect.signature` 探测    |
+| `[AutoML]` 完全没打                          | `modeling_step.py:170` 的 `if use_automl` 是否进了      |
+| AutoML 报 `optuna API 不兼容`               | `automl_trainer.py:236` 的 `try import` 块              |
+| 软取消无反应                                  | `worker_thread.py:run()` 末尾的 `cancel_event.is_set()` |
+| 二次确认窗没出来                              | `water_quality_gui.py:on_run_all_clicked` line ~2848    |
+| 9 步产物路径错位                              | `pipeline/runner.py:PIPELINE_STEPS` 的 `output` 字段    |
+| 老 v1 路径被走到                              | `_sync_legacy_attrs_from_context` 没调，或 v2 异常      |
+
+---
+
+> **作者注**：本清单对应**路线 B 一键全自动重构 4 部分全部落盘**的验收场景，编号与 todo 8 同步。
+> 跑通 §1 + §2 + §3 三段即视为 MVP 验收通过；§4 用于鲁棒性抽查。

license.lic

@@ -0,0 +1,8 @@
+{
+  "version": "1.0",
+  "product": "WaterQualityInversion",
+  "machine_code": "76E4992A5CF08BA570D6150908E04755",
+  "generated_at": "2026-05-28 14:21:35",
+  "expiry": "2099-12-31",
+  "signature": "DC9AB900D7033A281E54F41F3F76D026FFA75D635484D40C7F6FC1F6023E02AB"
+}

run_smoke.bat

@@ -0,0 +1,6 @@
+@echo off
+cd /d "D:\111\office\ZHLduijie\1.WQ\WQ_GUI"
+call venv\Scripts\activate.bat
+set PYTHONPATH=new\app\api;%PYTHONPATH%
+python -c "import _smoke_test_train; _smoke_test_train.test_load_train_df(); _smoke_test_train.test_get_model_pipeline_all_types(); _smoke_test_train.test_run_train_sync_linearregression_fast(); _smoke_test_train.test_run_train_sync_bad_csv(); _smoke_test_train.test_run_train_sync_bad_target(); print('OK')" > %TEMP%\smoke_log.txt 2>&1
+type %TEMP%\smoke_log.txt

src/core/algorithms/__init__.py

@@ -16,6 +16,15 @@ from src.core.algorithms.glint_detection.detectors import (
     remove_shoreline_buffer,
     calculate_glint_mask,
 )
+from src.core.algorithms.qaa.qaas_baseline import QAABaselineSolver
+from src.core.algorithms.concentration_inversion import (
+    ChlorophyllInversion,
+    CDOMInversion,
+    TurbidityInversion,
+    TotalNitrogenInversion,
+    TotalPhosphorusInversion,
+    ConcentrationPipeline,
+)
 
 __all__ = [
     # 插值
@@ -33,4 +42,13 @@ __all__ = [
     'create_shoreline_buffer',
     'remove_shoreline_buffer',
     'calculate_glint_mask',
+    # QAA
+    'QAABaselineSolver',
+    # 浓度反演
+    'ChlorophyllInversion',
+    'CDOMInversion',
+    'TurbidityInversion',
+    'TotalNitrogenInversion',
+    'TotalPhosphorusInversion',
+    'ConcentrationPipeline',
 ]

src/core/algorithms/concentration_inversion.py

@@ -0,0 +1,670 @@
+# -*- coding: utf-8 -*-
+"""
+水质浓度反演模块
+
+基于 QAA Step 8 输出的光谱吸收/散射系数 (a_lambda, bb_lambda)，
+通过生物光学模型反演水质参数浓度。
+
+主要反演目标：
+  - 叶绿素 A (Chl-a)：675nm 吸收峰法
+  - 浊度 (Turbidity)：后向散射系数法
+  - CDOM 吸收系数 a_dg(440)：指数衰减法
+  - 总氮 (TN) / 总磷 (TP)：光学代理回归框架
+
+参考：
+  - Lee, Z.P. et al. (2002/2010/2014) QAA 系列
+  - Bricaud, A. et al. (1998) Limnol. Oceanogr. — 叶绿素比吸收系数
+  - Carder, K.L. et al. (1999) Marine Technology Society — CDOM 指数衰减
+"""
+
+from __future__ import annotations
+
+import os
+from typing import Dict, List, Optional, Tuple, Union
+
+import numpy as np
+import pandas as pd
+
+
+# ------------------------------------------------------------------
+# 公共系数表（来自 Bricaud et al. 1998 等文献，内陆水体典型值）
+# ------------------------------------------------------------------
+
+# 叶绿素比吸收系数 a*_ph(675) 单位：m²/mg
+# 随叶绿素浓度范围变化，Bricaud 经验值
+CHLA_SPECIFIC_ABSORPTION: Dict[str, float] = {
+    "low":    0.055,   # 寡营养水体，Chla < 5 mg/m³
+    "medium": 0.040,   # 中营养，Chla 5-30 mg/m³
+    "high":   0.028,   # 富营养，Chla 30-100 mg/m³
+    "bloom":  0.020,   # 藻华，Chla > 100 mg/m³
+}
+
+# CDOM 指数衰减斜率 S（单位：nm⁻¹），内陆水体典型范围 0.010-0.025
+CDOM_S_LOOKUP: Dict[str, float] = {
+    "low_turbidity":   0.010,   # 清澈寡营养
+    "medium_turbidity": 0.015,  # 中等浊度
+    "high_turbidity":  0.020,   # 高浊度富营养
+    "bloom":           0.025,   # 藻华主导
+}
+
+# 纯水吸收系数表（400-800nm，Babin et al. 2003 简化值，单位：m⁻¹）
+PURE_WATER_A: Dict[int, float] = {
+    400: 0.0064, 410: 0.0066, 420: 0.0068, 430: 0.0072,
+    440: 0.0080, 450: 0.0092, 460: 0.0105, 470: 0.0120,
+    480: 0.0135, 490: 0.0155, 500: 0.0175, 510: 0.0200,
+    520: 0.0230, 530: 0.0270, 540: 0.0315, 550: 0.0370,
+    560: 0.0435, 570: 0.0510, 580: 0.0600, 590: 0.0710,
+    600: 0.0830, 610: 0.0960, 620: 0.1110, 630: 0.1280,
+    640: 0.1470, 650: 0.1680, 660: 0.1920, 670: 0.2180,
+    675: 0.2450, 680: 0.2750, 690: 0.3100, 700: 0.3500,
+    710: 0.3950, 720: 0.4450, 730: 0.5000, 740: 0.5600,
+    750: 0.6250, 760: 0.6950, 770: 0.7700, 780: 0.8500,
+    790: 0.9300, 800: 1.0100,
+}
+
+
+def _interp_pure_water_a(wavelength: float) -> float:
+    """线性插值获取纯水吸收系数"""
+    wl_int = {k for k in PURE_WATER_A if k <= int(wavelength)}
+    if not wl_int:
+        return PURE_WATER_A[min(PURE_WATER_A.keys())]
+    k_low = max(wl_int)
+    k_high = min({k for k in PURE_WATER_A if k >= int(wavelength)} or {k_low})
+    if k_low == k_high:
+        return float(PURE_WATER_A[k_low])
+    w = (wavelength - k_low) / (k_high - k_low)
+    return float(PURE_WATER_A[k_low]) * (1 - w) + float(PURE_WATER_A[k_high]) * w
+
+
+# ------------------------------------------------------------------
+# 叶绿素反演器
+# ------------------------------------------------------------------
+
+class ChlorophyllInversion:
+    """
+    基于 675nm 吸收峰法的叶绿素 A 浓度反演。
+
+    原理：
+        总吸收 a(675) = a_w(675) + a_ph(675) + a_dg(675)
+        其中 a_ph(675) 是叶绿素特征吸收峰，
+        a_dg(675) ≈ a_dg(440) * exp(-S * (675-440))
+
+    步骤：
+        1. 从 a(λ) 减去纯水吸收 a_w(λ)
+        2. 用线性基线法估算 a_dg(675)：baseline(675) = mean[a(665), a(685)]
+        3. a_ph(675) = a(675) - a_w(675) - baseline(675)
+        4. Chla = a_ph(675) / a*_ph(675)
+
+    Parameters
+    ----------
+    specific_absorption : float, optional
+        叶绿素比吸收系数 a*_ph(675)，单位 m²/mg。
+        若为 None，使用浓度自适应估算逻辑。
+    lake_case : str, optional
+        水体类型标识，用于自动选择比吸收系数，
+        支持 "oligotrophic_clear" / "medium" / "bloom_dominant" / "turbid_mixed"。
+    """
+
+    def __init__(
+        self,
+        specific_absorption: Optional[float] = None,
+        lake_case: Optional[str] = None
+    ):
+        self.specific_absorption = specific_absorption
+        self.lake_case = lake_case or "medium"
+
+    def run_inversion(
+        self,
+        wavelengths: np.ndarray,
+        a_lambda: np.ndarray,
+        bb_lambda: Optional[np.ndarray] = None
+    ) -> Dict:
+        """
+        执行叶绿素 A 反演。
+
+        Parameters
+        ----------
+        wavelengths : np.ndarray
+            波长数组（nm），形状 (n_bands,)。
+        a_lambda : np.ndarray
+            总吸收系数 a(λ)，形状 (n_bands,)。
+        bb_lambda : np.ndarray, optional
+            后向散射系数（暂未使用，保留扩展接口）。
+
+        Returns
+        -------
+        dict
+            包含键：
+              - chla_mg_m3 : 叶绿素 A 浓度（mg/m³）
+              - a_ph_675  : 675nm 处叶绿素吸收（m⁻¹）
+              - baseline_675 : 675nm 处 CDOM+NAP 基线（m⁻¹）
+              - a_w_675   : 纯水吸收（m⁻¹）
+        """
+        wavelengths = np.asarray(wavelengths, dtype=np.float64)
+        a_lambda = np.asarray(a_lambda, dtype=np.float64)
+
+        aw_675 = _interp_pure_water_a(675.0)
+
+        wl_arr = wavelengths
+        a_arr = a_lambda
+
+        a_665 = float(np.interp(665, wl_arr, a_arr, left=np.nan, right=np.nan))
+        a_675 = float(np.interp(675, wl_arr, a_arr, left=np.nan, right=np.nan))
+        a_685 = float(np.interp(685, wl_arr, a_arr, left=np.nan, right=np.nan))
+
+        if not np.isfinite(a_665) or not np.isfinite(a_675) or not np.isfinite(a_685):
+            return {
+                "chla_mg_m3": np.nan,
+                "a_ph_675": np.nan,
+                "baseline_675": np.nan,
+                "a_w_675": aw_675,
+                "warning": "675nm 波段缺失，无法进行叶绿素反演",
+            }
+
+        baseline_675 = (a_665 + a_685) / 2.0
+        a_ph_675 = max(a_675 - aw_675 - baseline_675, 0.0)
+
+        if self.specific_absorption is not None:
+            a_star = self.specific_absorption
+        else:
+            a_star = self._adaptive_specific_absorption(a_ph_675)
+
+        if a_star <= 0:
+            return {
+                "chla_mg_m3": np.nan,
+                "a_ph_675": a_ph_675,
+                "baseline_675": baseline_675,
+                "a_w_675": aw_675,
+                "warning": "比吸收系数为非正值",
+            }
+
+        chla = a_ph_675 / a_star
+        return {
+            "chla_mg_m3": chla,
+            "a_ph_675": a_ph_675,
+            "baseline_675": baseline_675,
+            "a_w_675": aw_675,
+        }
+
+    def _adaptive_specific_absorption(self, a_ph_675: float) -> float:
+        """根据 a_ph(675) 量级自适应选择比吸收系数"""
+        if a_ph_675 < 0.05:
+            return CHLA_SPECIFIC_ABSORPTION["low"]
+        elif a_ph_675 < 0.2:
+            return CHLA_SPECIFIC_ABSORPTION["medium"]
+        elif a_ph_675 < 0.5:
+            return CHLA_SPECIFIC_ABSORPTION["high"]
+        else:
+            return CHLA_SPECIFIC_ABSORPTION["bloom"]
+
+    def invert_to_csv(
+        self,
+        input_csv: str,
+        output_csv: str,
+        sample_id_col: str = "sample_id"
+    ) -> str:
+        """
+        从 a_lambda_results.csv 批量反演叶绿素并保存结果。
+
+        Parameters
+        ----------
+        input_csv : str
+            Step 8 输出的 a_lambda_results.csv 路径。
+        output_csv : str
+            保存路径。
+        sample_id_col : str
+            样本 ID 列名。
+
+        Returns
+        -------
+        str
+            输出文件路径。
+        """
+        df = pd.read_csv(input_csv, encoding="utf-8-sig")
+        df = df.sort_values([sample_id_col, "Wavelength"])
+
+        results = []
+        for sid, group in df.groupby(sample_id_col, sort=False):
+            wl = group["Wavelength"].values.astype(np.float64)
+            a = group["a_lambda"].values.astype(np.float64)
+            res = self.run_inversion(wl, a)
+            res[sample_id_col] = sid
+            results.append(res)
+
+        out_df = pd.DataFrame(results)
+        cols = [sample_id_col, "chla_mg_m3", "a_ph_675", "baseline_675", "a_w_675"]
+        cols = [c for c in cols if c in out_df.columns]
+        out_df = out_df[cols]
+        os.makedirs(os.path.dirname(output_csv) or ".", exist_ok=True)
+        out_df.to_csv(output_csv, index=False, float_format="%.6f")
+        return output_csv
+
+
+# ------------------------------------------------------------------
+# CDOM 反演器
+# ------------------------------------------------------------------
+
+class CDOMInversion:
+    """
+    基于指数衰减模型的 CDOM 吸收系数反演。
+
+    原理：
+        a_dg(λ) = a_dg(λ₀) * exp(-S * (λ - λ₀))
+
+    取 λ₀ = 440nm（蓝光峰），S 由水体类型决定，
+    通过 a(550) ≈ a_w(550) + a_dg(550) 反推 a_dg(440)。
+
+    Parameters
+    ----------
+    S : float, optional
+        CDOM 指数衰减斜率（nm⁻¹）。若为 None，根据 lake_case 自动选择。
+    reference_wavelength : int
+        参考波长，默认 440nm。
+    """
+
+    def __init__(
+        self,
+        S: Optional[float] = None,
+        reference_wavelength: int = 440
+    ):
+        self.S = S
+        self.ref_wl = reference_wavelength
+
+    def run_inversion(
+        self,
+        wavelengths: np.ndarray,
+        a_lambda: np.ndarray
+    ) -> Dict:
+        """
+        执行 CDOM 反演。
+
+        Parameters
+        ----------
+        wavelengths : np.ndarray
+            波长数组。
+        a_lambda : np.ndarray
+            总吸收系数 a(λ)。
+
+        Returns
+        -------
+        dict
+            包含键：
+              - a_dg_440 : 440nm 处 CDOM 吸收（m⁻¹）
+              - S        : 使用的衰减斜率
+        """
+        wavelengths = np.asarray(wavelengths, dtype=np.float64)
+        a_lambda = np.asarray(a_lambda, dtype=np.float64)
+
+        if self.S is None:
+            S = CDOM_S_LOOKUP["medium_turbidity"]
+        else:
+            S = self.S
+
+        a_440 = float(np.interp(440, wavelengths, a_lambda, left=np.nan, right=np.nan))
+        a_550 = float(np.interp(550, wavelengths, a_lambda, left=np.nan, right=np.nan))
+        aw_440 = _interp_pure_water_a(440.0)
+        aw_550 = _interp_pure_water_a(550.0)
+
+        a_dg_550 = max(a_550 - aw_550, 0.0)
+        delta_wl = 550 - self.ref_wl
+        a_dg_440 = a_dg_550 * np.exp(S * delta_wl)
+
+        return {
+            "a_dg_440": a_dg_440,
+            "a_dg_550": a_dg_550,
+            "S": S,
+        }
+
+    def invert_to_csv(
+        self,
+        input_csv: str,
+        output_csv: str,
+        sample_id_col: str = "sample_id"
+    ) -> str:
+        """从 a_lambda_results.csv 批量反演 CDOM 并保存结果。"""
+        df = pd.read_csv(input_csv, encoding="utf-8-sig")
+        df = df.sort_values([sample_id_col, "Wavelength"])
+
+        results = []
+        for sid, group in df.groupby(sample_id_col, sort=False):
+            wl = group["Wavelength"].values.astype(np.float64)
+            a = group["a_lambda"].values.astype(np.float64)
+            res = self.run_inversion(wl, a)
+            res[sample_id_col] = sid
+            results.append(res)
+
+        out_df = pd.DataFrame(results)
+        cols = [sample_id_col, "a_dg_440", "a_dg_550", "S"]
+        cols = [c for c in cols if c in out_df.columns]
+        out_df = out_df[cols]
+        os.makedirs(os.path.dirname(output_csv) or ".", exist_ok=True)
+        out_df.to_csv(output_csv, index=False, float_format="%.6f")
+        return output_csv
+
+
+# ------------------------------------------------------------------
+# 浊度反演器
+# ------------------------------------------------------------------
+
+class TurbidityInversion:
+    """
+    基于后向散射系数的光学浊度反演。
+
+    原理（简化模型）：
+        Turbidity (NTU) ≈ k * b_b(550)
+
+    其中 b_b(550) 是 550nm 处的后向散射系数，
+    k 为经验系数（内陆水体典型值 1.0-3.0）。
+
+    Parameters
+    ----------
+    k : float
+        经验系数。默认值 2.0。
+    reference_wavelength : int
+        参考波段，默认 550nm。
+    """
+
+    def __init__(self, k: float = 2.0, reference_wavelength: int = 550):
+        self.k = k
+        self.ref_wl = reference_wavelength
+
+    def run_inversion(
+        self,
+        wavelengths: np.ndarray,
+        bb_lambda: np.ndarray
+    ) -> Dict:
+        """
+        执行浊度反演。
+
+        Parameters
+        ----------
+        wavelengths : np.ndarray
+            波长数组。
+        bb_lambda : np.ndarray
+            后向散射系数 b_b(λ)。
+
+        Returns
+        -------
+        dict
+            包含键：
+              - turbidity_ntu : 浊度（NTU）
+              - bb_ref         : 参考波段处的 b_b 值
+        """
+        wavelengths = np.asarray(wavelengths, dtype=np.float64)
+        bb_lambda = np.asarray(bb_lambda, dtype=np.float64)
+
+        bb_ref = float(np.interp(
+            self.ref_wl, wavelengths, bb_lambda, left=np.nan, right=np.nan
+        ))
+        turbidity = self.k * bb_ref
+
+        return {
+            "turbidity_ntu": turbidity,
+            "bb_ref": bb_ref,
+        }
+
+    def invert_to_csv(
+        self,
+        input_csv: str,
+        output_csv: str,
+        sample_id_col: str = "sample_id"
+    ) -> str:
+        """从 a_lambda_results.csv 批量反演浊度并保存结果。"""
+        df = pd.read_csv(input_csv, encoding="utf-8-sig")
+        if "bb_lambda" not in df.columns:
+            raise ValueError("输入 CSV 中缺少 bb_lambda 列")
+        df = df.sort_values([sample_id_col, "Wavelength"])
+
+        results = []
+        for sid, group in df.groupby(sample_id_col, sort=False):
+            wl = group["Wavelength"].values.astype(np.float64)
+            bb = group["bb_lambda"].values.astype(np.float64)
+            res = self.run_inversion(wl, bb)
+            res[sample_id_col] = sid
+            results.append(res)
+
+        out_df = pd.DataFrame(results)
+        cols = [sample_id_col, "turbidity_ntu", "bb_ref"]
+        cols = [c for c in cols if c in out_df.columns]
+        out_df = out_df[cols]
+        os.makedirs(os.path.dirname(output_csv) or ".", exist_ok=True)
+        out_df.to_csv(output_csv, index=False, float_format="%.6f")
+        return output_csv
+
+
+# ------------------------------------------------------------------
+# 总氮 / 总磷反演器（光学代理回归框架）
+# ------------------------------------------------------------------
+
+class TotalNitrogenInversion:
+    """
+    总氮 (TN) 光学代理回归模型。
+
+    框架说明：
+        TN 与 Chla 之间通常存在正相关（R² ≈ 0.5-0.7），
+        本类提供回归框架，实际系数需由实测数据标定。
+
+    公式（线性代理）：
+        TN (mg/L) = α * Chla + β * Turbidity + γ
+
+    Parameters
+    ----------
+    alpha : float
+        Chla 系数。默认 0.05。
+    beta : float
+        浊度系数。默认 0.10。
+    gamma : float
+        截距。默认 0.20。
+    """
+
+    def __init__(
+        self,
+        alpha: float = 0.05,
+        beta: float = 0.10,
+        gamma: float = 0.20
+    ):
+        self.alpha = alpha
+        self.beta = beta
+        self.gamma = gamma
+
+    def run_inversion(
+        self,
+        chla_mg_m3: float,
+        turbidity_ntu: float
+    ) -> Dict:
+        """执行总氮反演（光学代理法）。"""
+        tn = self.alpha * chla_mg_m3 + self.beta * turbidity_ntu + self.gamma
+        return {"tn_mg_L": tn}
+
+    def calibrate(
+        self,
+        samples: List[Dict]
+    ) -> None:
+        """
+        用实测样本标定回归系数。
+
+        Parameters
+        ----------
+        samples : list[dict]
+            样本列表，每项包含 'chla', 'turbidity', 'tn' 键。
+        """
+        try:
+            import numpy as np
+            X = np.array([[s["chla"], s["turbidity"]] for s in samples])
+            y = np.array([s["tn"] for s in samples])
+            coeffs, _, _, _ = np.linalg.lstsq(X, y, rcond=None)
+            self.alpha, self.beta = coeffs
+            self.gamma = float(np.mean(y - self.alpha * X[:, 0] - self.beta * X[:, 1]))
+        except Exception as e:
+            raise RuntimeError(f"标定失败: {e}")
+
+
+class TotalPhosphorusInversion:
+    """
+    总磷 (TP) 光学代理回归模型。
+
+    框架说明：
+        TP 与 Chla / 浊度均相关（湖泊富营养化阶段尤为明显），
+        提供双变量线性回归框架，实际系数需由实测数据标定。
+
+    公式（线性代理）：
+        TP (mg/L) = α * Chla + β * Turbidity + γ
+
+    Parameters
+    ----------
+    alpha : float
+        Chla 系数。默认 0.002。
+    beta : float
+        浊度系数。默认 0.005。
+    gamma : float
+        截距。默认 0.010。
+    """
+
+    def __init__(
+        self,
+        alpha: float = 0.002,
+        beta: float = 0.005,
+        gamma: float = 0.010
+    ):
+        self.alpha = alpha
+        self.beta = beta
+        self.gamma = gamma
+
+    def run_inversion(
+        self,
+        chla_mg_m3: float,
+        turbidity_ntu: float
+    ) -> Dict:
+        """执行总磷反演（光学代理法）。"""
+        tp = self.alpha * chla_mg_m3 + self.beta * turbidity_ntu + self.gamma
+        return {"tp_mg_L": tp}
+
+    def calibrate(
+        self,
+        samples: List[Dict]
+    ) -> None:
+        """用实测样本标定回归系数。"""
+        try:
+            import numpy as np
+            X = np.array([[s["chla"], s["turbidity"]] for s in samples])
+            y = np.array([s["tp"] for s in samples])
+            coeffs, _, _, _ = np.linalg.lstsq(X, y, rcond=None)
+            self.alpha, self.beta = coeffs
+            self.gamma = float(np.mean(y - self.alpha * X[:, 0] - self.beta * X[:, 1]))
+        except Exception as e:
+            raise RuntimeError(f"标定失败: {e}")
+
+
+# ------------------------------------------------------------------
+# 一站式浓度反演流水线
+# ------------------------------------------------------------------
+
+class ConcentrationPipeline:
+    """
+    整合 Chlorophyll / CDOM / Turbidity / TN / TP 反演的一站式流水线。
+
+    接收 Step 8 输出的 a_lambda_results.csv，
+    输出 final_concentrations.csv（含所有水质参数浓度列）。
+
+    Parameters
+    ----------
+    lake_case : str, optional
+        水体类型，用于 Chla 比吸收系数自适应选择。
+    S_cdom : float, optional
+        CDOM 衰减斜率（若为 None，自动选择）。
+    k_turbidity : float
+        浊度经验系数。
+    tn_params : dict, optional
+        总氮反演初始参数。
+    tp_params : dict, optional
+        总磷反演初始参数。
+    """
+
+    def __init__(
+        self,
+        lake_case: str = "medium",
+        S_cdom: Optional[float] = None,
+        k_turbidity: float = 2.0,
+        tn_params: Optional[Dict] = None,
+        tp_params: Optional[Dict] = None,
+    ):
+        self.lake_case = lake_case
+        self.chla_inv = ChlorophyllInversion(lake_case=lake_case)
+        self.cdom_inv = CDOMInversion(S=S_cdom)
+        self.turb_inv = TurbidityInversion(k=k_turbidity)
+        self.tn_inv = TotalNitrogenInversion(**(tn_params or {}))
+        self.tp_inv = TotalPhosphorusInversion(**(tp_params or {}))
+
+    def run_pipeline(
+        self,
+        input_csv: str,
+        output_csv: str,
+        sample_id_col: str = "sample_id"
+    ) -> str:
+        """
+        执行完整浓度反演流水线。
+
+        Parameters
+        ----------
+        input_csv : str
+            Step 8 输出的 a_lambda_results.csv 路径。
+        output_csv : str
+            输出 final_concentrations.csv 路径。
+        sample_id_col : str
+            样本 ID 列名。
+
+        Returns
+        -------
+        str
+            输出文件路径。
+        """
+        df = pd.read_csv(input_csv, encoding="utf-8-sig")
+        if "bb_lambda" not in df.columns:
+            df["bb_lambda"] = np.nan
+
+        # ── 保留原始坐标列：按 sample_id 取第一条记录的非光谱列 ───────────
+        wl_col = "Wavelength"
+        coord_meta_cols = [c for c in df.columns if c not in (sample_id_col, wl_col, "a_lambda", "bb_lambda")]
+        coord_df = df.groupby(sample_id_col, sort=False)[coord_meta_cols].first().reset_index()
+
+        df = df.sort_values([sample_id_col, "Wavelength"])
+
+        results = []
+        for sid, group in df.groupby(sample_id_col, sort=False):
+            wl = group["Wavelength"].values.astype(np.float64)
+            a = group["a_lambda"].values.astype(np.float64)
+            bb = group["bb_lambda"].values.astype(np.float64) \
+                if "bb_lambda" in group.columns and group["bb_lambda"].notna().any() \
+                else None
+
+            chla_res = self.chla_inv.run_inversion(wl, a)
+            cdom_res = self.cdom_inv.run_inversion(wl, a)
+            if bb is not None and np.any(np.isfinite(bb)):
+                turb_res = self.turb_inv.run_inversion(wl, bb)
+            else:
+                turb_res = {"turbidity_ntu": np.nan, "bb_ref": np.nan}
+
+            chla_val = chla_res.get("chla_mg_m3", np.nan)
+            turb_val = turb_res.get("turbidity_ntu", np.nan)
+
+            tn_res = self.tn_inv.run_inversion(chla_val, turb_val)
+            tp_res = self.tp_inv.run_inversion(chla_val, turb_val)
+
+            row = {
+                sample_id_col: sid,
+                "Chla_mg_m3": chla_val,
+                "a_ph_675_m1": chla_res.get("a_ph_675", np.nan),
+                "CDOM_a_dg_440_m1": cdom_res.get("a_dg_440", np.nan),
+                "Turbidity_NTU": turb_val,
+                "TN_mg_L": tn_res.get("tn_mg_L", np.nan),
+                "TP_mg_L": tp_res.get("tp_mg_L", np.nan),
+            }
+            results.append(row)
+
+        out_df = pd.DataFrame(results)
+        # ── 将原始坐标列按 sample_id 合并到浓度结果左侧 ───────────────────
+        if not coord_df.empty and sample_id_col in coord_df.columns:
+            out_df = coord_df.merge(out_df, on=sample_id_col, how="left")
+        os.makedirs(os.path.dirname(output_csv) or ".", exist_ok=True)
+        out_df.to_csv(output_csv, index=False, float_format="%.6f")
+        return output_csv

src/core/algorithms/interpolation/interpolator.py

@@ -3,8 +3,24 @@
 
 提供对影像中所有波段都为0的像素点进行插值的核心数学逻辑。
 支持多种插值方法：nearest, bilinear, spline (RBF), kriging。
+
+本模块使用多进程并行分块 IO 加速（Plan A）：
+- ProcessPoolExecutor 为每个 worker 进程打开一次源影像（initializer 阶段），
+  避免每块重复 gdal.Open 带来的开销（Windows 上 ~50ms/次）
+- 主进程统一负责输出文件的写入，避免多进程写锁竞争
+- 分块大小（block_size）默认 1024，内存充足可调至 2048 / 4096
+
+注意：
+- GDAL Dataset / Rasterio Dataset 对象不能跨进程传递（picking 不支持），
+  所以 worker 必须在 init 阶段自己独立打开源文件
+- 每个 worker 强制设置 ``GDAL_NUM_THREADS=1``，避免 8 worker × GDAL 多线程
+  造成的 CPU 过订阅
+- 关闭多进程：传 ``use_multiprocessing=False`` 或 ``n_workers=1``
 """
 
+import multiprocessing
+from concurrent.futures import ProcessPoolExecutor
+
 import numpy as np
 from typing import Optional, Union, Tuple, List
 from pathlib import Path
@@ -24,6 +40,9 @@ except ImportError:
     GDAL_AVAILABLE = False
 
 
+_worker_dataset: Optional["gdal.Dataset"] = None
+
+
 def interpolate_pixels(
     image_stack: np.ndarray,
     zero_coords: np.ndarray,
@@ -52,7 +71,6 @@ def interpolate_pixels(
     height, width, n_bands = image_stack.shape
     result = image_stack.copy()
 
-    # 兼容中文和各种格式的method参数
     raw_method = str(interpolation_method).lower()
     if 'nearest' in raw_method or '邻近' in raw_method or '最邻近' in raw_method:
         method = 'nearest'
@@ -181,39 +199,271 @@ def _interpolate_single_band(
     return np.zeros(len(zero_coords))
 
 
+def _normalize_interpolation_method(method: str) -> str:
+    """将中文/英文混用的插值方法名归一化为内部标准名
+
+    支持: 'nearest'/'邻近'/'最邻近'，'bilinear'/'线性'/'双线性'，
+    'spline'/'样条'/'rbf'，'kriging'/'克里金'。
+    """
+    raw = str(method).lower()
+    if 'nearest' in raw or '邻近' in raw or '最邻近' in raw:
+        return 'nearest'
+    if 'bilinear' in raw or '线性' in raw or '双线性' in raw:
+        return 'bilinear'
+    if 'spline' in raw or '样条' in raw or 'rbf' in raw:
+        return 'spline'
+    if 'kriging' in raw or '克里金' in raw:
+        return 'kriging'
+    return 'nearest'
+
+
+def _read_water_mask_to_array(
+    water_mask: Optional[Union[str, np.ndarray]],
+    expected_height: int,
+    expected_width: int,
+) -> Optional[np.ndarray]:
+    """读取水域掩膜为 numpy 数组（单波段，bool/int 均可）
+
+    None 或空字符串直接返回 None。形状不匹配时给出告警但不抛错，
+    让调用方按"无掩膜"路径继续。
+    """
+    if water_mask is None:
+        return None
+    if isinstance(water_mask, str):
+        if not water_mask.strip():
+            return None
+        mask_ds = gdal.Open(water_mask, gdal.GA_ReadOnly)
+        if mask_ds is None:
+            print(f"  [warn] 无法打开水域掩膜 {water_mask}，按无掩膜处理")
+            return None
+        try:
+            mask_array = mask_ds.GetRasterBand(1).ReadAsArray()
+        finally:
+            mask_ds = None
+    elif isinstance(water_mask, np.ndarray):
+        mask_array = water_mask
+    else:
+        return None
+
+    if mask_array.shape != (expected_height, expected_width):
+        print(
+            f"  [warn] 水域掩膜形状 {mask_array.shape} 与影像 "
+            f"({expected_height}, {expected_width}) 不匹配，按无掩膜处理"
+        )
+        return None
+    return mask_array
+
+
+def _init_worker(img_path: str) -> None:
+    """ProcessPoolExecutor initializer: 每个 worker 进程只调用一次
+
+    在 worker 进程启动时打开源影像 dataset 并缓存在模块全局变量
+    ``_worker_dataset`` 中。后续所有块处理直接复用这个 dataset，
+    避免每块重复 ``gdal.Open``（Windows 上约 50ms/次，100 块即 5s）。
+
+    同时设置 ``GDAL_NUM_THREADS=1``，避免 8 worker × GDAL 默认多线程
+    造成的 CPU 过订阅。
+    """
+    global _worker_dataset
+    gdal.SetConfigOption('GDAL_NUM_THREADS', '1')
+    if hasattr(gdal, 'UseExceptions'):
+        gdal.UseExceptions()
+    _worker_dataset = gdal.Open(img_path, gdal.GA_ReadOnly)
+    if _worker_dataset is None:
+        raise RuntimeError(f"Worker failed to open source image: {img_path}")
+
+
+def _interpolate_block_worker(task: tuple) -> tuple:
+    """ProcessPoolExecutor worker: 处理单个块并返回结果
+
+    该函数必须保持模块级（可被 pickle），不持有任何外部状态——
+    源 dataset 通过 ``_worker_dataset`` 模块全局变量获取。
+
+    Returns:
+        ``(x0, y0, inner_bands, zero_count, error_msg)`` 元组：
+        - x0, y0: 块在影像中的写入起点
+        - inner_bands: ``List[np.ndarray]``，每个元素是 (inner_h, inner_w)
+                       float32 数组（每个波段一个），或失败时为 None
+        - zero_count: 该扩展块中识别到的零像素数（含 halo 范围）
+        - error_msg: None 表示成功，str 表示错误信息
+    """
+    (
+        x0, y0, ey0, ex0, ey1, ex1,
+        row_offset, col_offset, inner_h, inner_w,
+        mask_segment_ext, method,
+    ) = task
+    if _worker_dataset is None:
+        return (x0, y0, None, 0, "Worker dataset not initialized")
+    try:
+        inner_bands, zero_count = _process_one_block(
+            _worker_dataset, x0, y0, ey0, ex0, ey1, ex1,
+            row_offset, col_offset, inner_h, inner_w,
+            mask_segment_ext, method,
+        )
+        return (x0, y0, inner_bands, zero_count, None)
+    except Exception as e:
+        return (x0, y0, None, 0, str(e))
+
+
+def _process_one_block(
+    dataset: "gdal.Dataset",
+    x0: int, y0: int,
+    ey0: int, ex0: int, ey1: int, ex1: int,
+    row_offset: int, col_offset: int,
+    inner_h: int, inner_w: int,
+    mask_segment_ext: Optional[np.ndarray],
+    method: str,
+) -> Tuple[List[np.ndarray], int]:
+    """处理单个扩展块（纯计算核心，dataset 显式传入）
+
+    串行模式和并行模式共用此函数。并行模式下 dataset 来自 worker 的
+    缓存（``_worker_dataset``），串行模式下 dataset 由主函数传入。
+
+    Args:
+        dataset: 已打开的源影像 dataset
+        x0, y0: 内部块左上角（写入位置）
+        ey0, ex0, ey1, ex1: 扩展块（含 halo）坐标
+        row_offset, col_offset: 内部块在扩展块中的偏移
+        inner_h, inner_w: 内部块尺寸
+        mask_segment_ext: 扩展块对应的水域掩膜（None 表示不应用）
+        method: 插值方法（已归一化）
+
+    Returns:
+        ``(inner_bands, zero_count)`` 元组：
+        - inner_bands: ``List[np.ndarray]``，长度 = n_bands，每个元素形状为
+                       ``(inner_h, inner_w)`` 的 float32 数组
+        - zero_count: 扩展块中识别到的零像素数
+    """
+    n_bands = dataset.RasterCount
+    ext_bands: List[np.ndarray] = []
+    for b in range(1, n_bands + 1):
+        band = dataset.GetRasterBand(b)
+        ext_bands.append(
+            band.ReadAsArray(ex0, ey0, ex1 - ex0, ey1 - ey0).astype(np.float32)
+        )
+        band = None
+
+    try:
+        ext_h, ext_w = ey1 - ey0, ex1 - ex0
+
+        all_zero_ext = np.ones((ext_h, ext_w), dtype=bool)
+        for b_data in ext_bands:
+            all_zero_ext &= (b_data == 0)
+
+        if mask_segment_ext is not None:
+            all_zero_ext &= (mask_segment_ext > 0)
+
+        zero_count = int(np.sum(all_zero_ext))
+
+        if zero_count == 0:
+            inner_bands = [
+                ext_bands[b][
+                    row_offset:row_offset + inner_h,
+                    col_offset:col_offset + inner_w,
+                ]
+                for b in range(n_bands)
+            ]
+            return inner_bands, 0
+
+        zero_y, zero_x = np.where(all_zero_ext)
+        zero_coords = np.column_stack([zero_x, zero_y])
+
+        valid_mask = ~all_zero_ext
+        valid_y, valid_x = np.where(valid_mask)
+        valid_coords = np.column_stack([valid_x, valid_y])
+
+        if len(valid_coords) == 0:
+            print(
+                f"  [warn] 块 (y={y0}-{y0 + inner_h}, x={x0}-{x0 + inner_w}) "
+                f"无有效像素可作插值上下文，已跳过"
+            )
+            inner_bands = [
+                ext_bands[b][
+                    row_offset:row_offset + inner_h,
+                    col_offset:col_offset + inner_w,
+                ]
+                for b in range(n_bands)
+            ]
+            return inner_bands, zero_count
+
+        for b in range(n_bands):
+            ext_band = ext_bands[b]
+            valid_values_band = ext_band[valid_mask]
+            if len(valid_values_band) == 0:
+                continue
+            band_result = _interpolate_single_band(
+                zero_coords, valid_coords, valid_values_band, method
+            )
+            ext_band[zero_y, zero_x] = band_result
+
+        inner_bands = [
+            ext_bands[b][
+                row_offset:row_offset + inner_h,
+                col_offset:col_offset + inner_w,
+            ]
+            for b in range(n_bands)
+        ]
+        return inner_bands, zero_count
+    finally:
+        del ext_bands
+
+
 def interpolate_zero_pixels_batch(
     img_path: str,
     interpolation_method: str = 'nearest',
     output_path: Optional[str] = None,
     water_mask: Optional[Union[str, np.ndarray]] = None,
     deglint_dir: Optional[str] = None,
-    callback_progress: Optional[callable] = None
+    callback_progress: Optional[callable] = None,
+    block_size: int = 1024,
+    halo_size: int = 64,
+    n_workers: Optional[int] = None,
+    use_multiprocessing: bool = True,
 ) -> Tuple[str, Optional[np.ndarray]]:
     """
-    对影像中所有波段都为0的像素点进行插值（完整流程，含文件I/O）
+    对影像中所有波段都为0的像素点进行插值（完整流程，含文件I/O）。
+
+    采用 **分块 IO + 多进程并行** 策略：
+    1. 影像按 ``block_size`` × ``block_size`` 分块，每块边界外扩展
+       ``halo_size`` 像素作为插值上下文，避免块边缘插值退化
+    2. 多进程并行（默认 ``ProcessPoolExecutor``，worker 数 = CPU 核心数）
+       并发处理所有块；GDAL Dataset 不能跨进程传递，所以每个 worker
+       在 ``initializer`` 阶段独立打开源文件一次并缓存
+    3. 主进程按块序接收处理结果并统一写入输出文件，避免写锁竞争
+    4. 该方案可彻底避免一次性读取 50 波段整景影像时的 OOM 隐患
+       （50 波段 × 4000×4000 × float32 ≈ 3GB 的 np.dstack）
 
     Args:
         img_path: 输入影像文件路径
-        interpolation_method: 插值方法，支持 'nearest', 'bilinear', 'spline', 'kriging'
-        output_path: 输出文件路径（如果为None，自动生成）
-        water_mask: 水域掩膜（文件路径或数组）
+        interpolation_method: 插值方法，支持 'nearest', 'bilinear', 'spline',
+            'kriging' 及其中文别名（'邻近'/'最邻近'/'线性'/'双线性'/'样条'/'克里金'）
+        output_path: 输出文件路径（如果为 None 且 deglint_dir 提供，自动生成）
+        water_mask: 水域掩膜（文件路径或数组），形状须与影像高宽一致
         deglint_dir: 去耀斑目录（用于生成默认输出路径）
-        callback_progress: 进度回调函数
+        callback_progress: 进度回调函数，签名 ``callback(msg: str)``
+        block_size: 分块大小（像素），默认 1024；内存充足可调 2048/4096
+        halo_size: 上下文 halo 宽度（像素），默认 64
+        n_workers: 并行 worker 进程数；None = ``multiprocessing.cpu_count()``；
+            传 1 等价于串行模式
+        use_multiprocessing: 是否启用多进程；False 时强制串行
 
     Returns:
-        (output_path, interpolated_image_stack) 元组
+        ``(output_path, None)`` 元组。第二个值固定为 ``None``（与原版语义保留
+        兼容；返回完整内存堆叠会重新引入 OOM 风险，故不再提供）。
     """
     if not SCIPY_AVAILABLE:
         raise ImportError("scipy未安装，无法进行0值像素插值")
     if not GDAL_AVAILABLE:
         raise ImportError("GDAL未安装，无法读取影像文件")
 
-    # 确定输出路径
-    if output_path is None and deglint_dir is not None:
-        output_path = str(Path(deglint_dir) / f"interpolated_{interpolation_method}.bsq")
+    method = _normalize_interpolation_method(interpolation_method)
 
-    # 检查文件是否已存在
-    if output_path and Path(output_path).exists():
+    if output_path is None and deglint_dir is not None:
+        output_path = str(Path(deglint_dir) / f"interpolated_{method}.bsq")
+    if output_path is None:
+        raise ValueError("output_path 和 deglint_dir 至少需要指定一个")
+
+    if Path(output_path).exists():
         return output_path, None
 
     dataset = gdal.Open(img_path, gdal.GA_ReadOnly)
@@ -227,94 +477,126 @@ def interpolate_zero_pixels_batch(
         geotransform = dataset.GetGeoTransform()
         projection = dataset.GetProjection()
 
-        # 读取所有波段数据
-        all_bands = []
-        for band_idx in range(1, n_bands + 1):
-            band = dataset.GetRasterBand(band_idx)
-            band_data = band.ReadAsArray().astype(np.float32)
-            all_bands.append(band_data)
-
-        image_stack = np.dstack(all_bands)
-
-        # 读取水域掩膜
-        mask_array = None
-        if water_mask is not None:
-            if isinstance(water_mask, str):
-                mask_dataset = gdal.Open(water_mask, gdal.GA_ReadOnly)
-                if mask_dataset:
-                    mask_array = mask_dataset.GetRasterBand(1).ReadAsArray()
-                    mask_dataset = None
-            elif isinstance(water_mask, np.ndarray):
-                mask_array = water_mask
-
-        # 找出所有波段都为0的像素点
-        all_bands_zero = np.all(image_stack == 0, axis=2)
-
-        if mask_array is not None:
-            all_bands_zero = all_bands_zero & (mask_array > 0)
-
-        zero_pixel_count = np.sum(all_bands_zero)
-        if zero_pixel_count == 0:
-            # 无需插值，直接保存
-            if output_path:
-                driver = gdal.GetDriverByName('ENVI')
-                if driver is None:
-                    driver = gdal.GetDriverByName('GTiff')
-                out_dataset = driver.Create(output_path, width, height, n_bands, gdal.GDT_Float32)
-                out_dataset.SetGeoTransform(geotransform)
-                out_dataset.SetProjection(projection)
-                for i, band_data in enumerate(all_bands):
-                    out_band = out_dataset.GetRasterBand(i + 1)
-                    out_band.WriteArray(band_data)
-                    out_band.FlushCache()
-                out_dataset = None
-            return output_path, image_stack
-
-        # 获取坐标
-        zero_y, zero_x = np.where(all_bands_zero)
-        zero_coords = np.column_stack([zero_x, zero_y])
-
-        valid_mask = ~all_bands_zero
-        valid_y, valid_x = np.where(valid_mask)
-        valid_coords = np.column_stack([valid_x, valid_y])
-
-        if len(valid_coords) == 0:
-            raise ValueError("没有有效像素可用于插值")
-
-        # 逐波段插值
-        interpolated_bands = []
-        for band_idx in range(n_bands):
-            if callback_progress:
-                callback_progress(f"处理波段 {band_idx + 1}/{n_bands}...")
-            band_data = all_bands[band_idx].copy()
-            valid_values_band = band_data[valid_mask]
-
-            if len(valid_values_band) == 0:
-                interpolated_bands.append(band_data)
-                continue
-
-            band_result = _interpolate_single_band(
-                zero_coords, valid_coords, valid_values_band, interpolation_method
+        if width <= 0 or height <= 0 or n_bands <= 0:
+            raise ValueError(
+                f"影像尺寸异常: width={width}, height={height}, n_bands={n_bands}"
             )
-            band_data[all_bands_zero] = band_result
-            interpolated_bands.append(band_data)
 
-        # 保存结果
-        if output_path:
-            driver = gdal.GetDriverByName('ENVI')
-            if driver is None:
-                driver = gdal.GetDriverByName('GTiff')
-            out_dataset = driver.Create(output_path, width, height, n_bands, gdal.GDT_Float32)
-            out_dataset.SetGeoTransform(geotransform)
-            out_dataset.SetProjection(projection)
-            for i, band_data in enumerate(interpolated_bands):
-                out_band = out_dataset.GetRasterBand(i + 1)
-                out_band.WriteArray(band_data)
-                out_band.FlushCache()
+        mask_array = _read_water_mask_to_array(water_mask, height, width)
+
+        driver = gdal.GetDriverByName('ENVI')
+        if driver is None:
+            driver = gdal.GetDriverByName('GTiff')
+        if driver is None:
+            raise RuntimeError("未找到可用的栅格驱动（ENVI / GTiff 都不存在）")
+
+        out_dataset = driver.Create(
+            output_path, width, height, n_bands, gdal.GDT_Float32
+        )
+        if out_dataset is None:
+            raise RuntimeError(f"无法创建输出文件: {output_path}")
+        out_dataset.SetGeoTransform(geotransform)
+        out_dataset.SetProjection(projection)
+
+        try:
+            if not use_multiprocessing:
+                effective_workers = 1
+            elif n_workers is not None and n_workers >= 1:
+                effective_workers = int(n_workers)
+            else:
+                try:
+                    cpu_count = multiprocessing.cpu_count() or 1
+                except (NotImplementedError, OSError):
+                    cpu_count = 1
+                # 为了内存安全，强制将物理进程数限制在最高 6 个
+                effective_workers = min(6, max(1, cpu_count))
+
+            n_blocks_y = (height + block_size - 1) // block_size
+            n_blocks_x = (width + block_size - 1) // block_size
+            total_blocks = n_blocks_y * n_blocks_x
+
+            tasks = []
+            for by in range(n_blocks_y):
+                y0 = by * block_size
+                y1 = min(y0 + block_size, height)
+                inner_h = y1 - y0
+                ey0 = max(0, y0 - halo_size)
+                ey1 = min(height, y1 + halo_size)
+                for bx in range(n_blocks_x):
+                    x0 = bx * block_size
+                    x1 = min(x0 + block_size, width)
+                    inner_w = x1 - x0
+                    ex0 = max(0, x0 - halo_size)
+                    ex1 = min(width, x1 + halo_size)
+                    row_offset = y0 - ey0
+                    col_offset = x0 - ex0
+                    mask_segment_ext = None
+                    if mask_array is not None:
+                        mask_segment_ext = mask_array[ey0:ey1, ex0:ex1].copy()
+                    tasks.append((
+                        x0, y0, ey0, ex0, ey1, ex1,
+                        row_offset, col_offset, inner_h, inner_w,
+                        mask_segment_ext, method,
+                    ))
+
+            if callback_progress:
+                callback_progress(
+                    f"分块插值开始: 共 {total_blocks} 块 "
+                    f"(block_size={block_size}, halo={halo_size}, method={method}, "
+                    f"workers={effective_workers})"
+                )
+
+            total_zero_pixels = 0
+
+            if effective_workers <= 1:
+                for block_idx, task in enumerate(tasks, 1):
+                    x0_t, y0_t = task[0], task[1]
+                    if callback_progress:
+                        callback_progress(
+                            f"块 {block_idx}/{total_blocks} "
+                            f"y=[{y0_t},{y0_t + task[8]}) x=[{x0_t},{x0_t + task[9]})"
+                        )
+                    inner_bands, zero_count = _process_one_block(
+                        dataset, *task
+                    )
+                    for b_idx, band_data in enumerate(inner_bands):
+                        out_dataset.GetRasterBand(b_idx + 1).WriteArray(
+                            band_data, xoff=x0_t, yoff=y0_t
+                        )
+                    total_zero_pixels += zero_count
+            else:
+                with ProcessPoolExecutor(
+                    max_workers=effective_workers,
+                    initializer=_init_worker,
+                    initargs=(img_path,),
+                ) as executor:
+                    futures = [
+                        executor.submit(_interpolate_block_worker, task)
+                        for task in tasks
+                    ]
+                    for block_idx, future in enumerate(futures, 1):
+                        x0_t, y0_t, inner_bands, zero_count, error = future.result()
+                        if error is not None:
+                            raise RuntimeError(
+                                f"块 (y={y0_t}, x={x0_t}) 处理失败: {error}"
+                            )
+                        if inner_bands is not None:
+                            for b_idx, band_data in enumerate(inner_bands):
+                                out_dataset.GetRasterBand(b_idx + 1).WriteArray(
+                                    band_data, xoff=x0_t, yoff=y0_t
+                                )
+                        total_zero_pixels += zero_count
+                        if callback_progress:
+                            callback_progress(f"已写入块 {block_idx}/{total_blocks}")
+
+            if callback_progress:
+                callback_progress(
+                    f"分块插值完成: 共处理 {total_zero_pixels} 个零像素 "
+                    f"（{total_blocks} 块，方法 {method}，workers={effective_workers}）"
+                )
+
+            return output_path, None
+        finally:
             out_dataset = None
-
-        result_stack = np.dstack(interpolated_bands)
-        return output_path, result_stack
-
     finally:
         dataset = None

src/core/algorithms/qaa/__init__.py

@@ -0,0 +1,7 @@
+# -*- coding: utf-8 -*-
+"""
+QAA 准解析反演算法模块
+"""
+from src.core.algorithms.qaa.qaas_baseline import QAABaselineSolver
+
+__all__ = ['QAABaselineSolver']

src/core/algorithms/qaa/qaas_baseline.py

@@ -0,0 +1,345 @@
+# -*- coding: utf-8 -*-
+"""
+QAA 准解析算法基线求解器 (QAABaselineSolver)
+
+实现 QAA-v5 / QAA-v6 核心步骤：
+  1. Rrs(λ) → r_rs(λ)（水面以下遥感反射率转换）
+  2. 计算中间变量 u(λ)（固有光学性质比值）
+  3. λ₀ 锚点查表获取纯水吸收 aw(λ₀) 和后向散射 bbw(λ₀)
+  4. 估算全波段 b_b(λ)（后向散射系数）
+  5. 推导全波段 a(λ)（总吸收系数）
+
+参考：
+  - Lee, Z.P. et al. (2002) JGR-Oceans, 107(C4), 9-1~9-18 (QAA-v4)
+  - Lee, Z.P. et al. (2010) Applied Optics, 49(4), 617-623 (QAA-v5)
+  - Lee, Z.P. et al. (2014) Applied Optics, 53(4), 598-611 (QAA-v6)
+"""
+
+import os
+import warnings
+from typing import Optional, Union, Tuple
+
+import numpy as np
+import pandas as pd
+
+
+class QAABaselineSolver:
+    """
+    QAA 准解析算法基线求解器。
+
+    Parameters
+    ----------
+    pure_water_csv : str, optional
+        纯水 IOPs 表路径，默认使用 src/utils/pure_water_iops.csv。
+    qaa_version : str, default "QAA-v6"
+        算法版本，支持 "QAA-v5" 或 "QAA-v6"。
+
+    Attributes
+    ----------
+    iops_df : pd.DataFrame
+        纯水 IOPs 表，含 Wavelength / aw / bbw 三列。
+    """
+
+    def __init__(
+        self,
+        pure_water_csv: Optional[str] = None,
+        qaa_version: str = "QAA-v6"
+    ):
+        if pure_water_csv is None:
+            project_root = os.path.abspath(
+                os.path.join(os.path.dirname(__file__), '..', '..', '..', 'utils')
+            )
+            pure_water_csv = os.path.join(project_root, 'pure_water_iops.csv')
+
+        if not os.path.exists(pure_water_csv):
+            raise FileNotFoundError(f"纯水 IOPs 表不存在: {pure_water_csv}")
+
+        self.iops_df = pd.read_csv(pure_water_csv)
+        self.qaa_version = qaa_version
+
+    # ------------------------------------------------------------------
+    # 核心 QAA 步骤
+    # ------------------------------------------------------------------
+
+    @staticmethod
+    def _rrs_to_rrs_subsurface(rrs: np.ndarray) -> np.ndarray:
+        """
+        将水面遥感反射率 Rrs 转换为水面以下遥感反射率 r_rs。
+
+        转换公式（Lee et al. 1999）：
+            r_rs = Rrs / (0.52 + 1.7 * Rrs)
+
+        Parameters
+        ----------
+        rrs : np.ndarray
+            水面遥感反射率 Rrs，形状 (N,) 或 (N, n_bands)。
+
+        Returns
+        -------
+        np.ndarray
+            水面以下遥感反射率 r_rs。
+        """
+        rrs = np.asarray(rrs, dtype=np.float64)
+        denom = 0.52 + 1.7 * rrs
+        with np.errstate(divide='ignore', invalid='ignore'):
+            result = rrs / denom
+        result[~np.isfinite(result)] = np.nan
+        return result
+
+    @staticmethod
+    def _compute_u(rrs_subsurface: np.ndarray) -> np.ndarray:
+        """
+        计算中间变量 u = b_b / (a + b_b)。
+
+        QAA-v5/v6 经验关系（Lee et al. 2002）：
+            u = r_rs / (0.5 * r_rs + sqrt(0.25 * r_rs^2 + 0.1 * r_rs))
+
+        Parameters
+        ----------
+        rrs_subsurface : np.ndarray
+            水面以下遥感反射率 r_rs。
+
+        Returns
+        -------
+        np.ndarray
+            u 值，范围 [0, 1)。
+        """
+        rs = np.asarray(rrs_subsurface, dtype=np.float64)
+        with np.errstate(divide='ignore', invalid='ignore'):
+            result = rs / (0.5 * rs + np.sqrt(0.25 * rs ** 2 + 0.1 * rs))
+        result[~np.isfinite(result)] = np.nan
+        return result
+
+    def _get_pure_water_iops(self, wavelength: Union[int, float]) -> Tuple[float, float]:
+        """
+        根据波长从纯水 IOPs 表中插值获取 aw 和 bbw。
+
+        Parameters
+        ----------
+        wavelength : float
+            波长（nm），范围应在 400-800nm 内。
+
+        Returns
+        -------
+        (aw, bbw) : tuple
+            纯水吸收系数 (m^-1) 和后向散射系数 (m^-1)。
+        """
+        df = self.iops_df
+        wl_arr = df['Wavelength'].values
+        aw_arr = df['aw'].values
+        bbw_arr = df['bbw'].values
+
+        aw = float(np.interp(wavelength, wl_arr, aw_arr))
+        bbw = float(np.interp(wavelength, wl_arr, bbw_arr))
+        return aw, bbw
+
+    @staticmethod
+    def _compute_bb(
+        u: np.ndarray,
+        bbw_0: float,
+        wavelength: np.ndarray,
+        lambda_0: int
+    ) -> np.ndarray:
+        """
+        估算全波段后向散射系数 b_b(λ)。
+
+        经验光谱形状（Lee et al. 2002, QAA-v4）：
+            b_b(λ) = b_bw(λ₀) * (λ₀ / λ)^S
+
+        其中 S 为经验光谱斜率参数（QAA-v5 中默认 0.5，
+        QAA-v6 中随 λ₀ 自适应调整）。
+
+        Parameters
+        ----------
+        u : np.ndarray
+            中间变量 u。
+        bbw_0 : float
+            λ₀ 处的纯水后向散射系数。
+        wavelength : np.ndarray
+            全波段波长数组。
+        lambda_0 : int
+            参考波长（锚点）。
+
+        Returns
+        -------
+        np.ndarray
+            全波段后向散射系数 b_b。
+        """
+        S = 0.5 if lambda_0 < 600 else 0.0
+        wavelength = np.asarray(wavelength, dtype=np.float64)
+        ratio = (float(lambda_0) / wavelength) ** S
+        bb = u * bbw_0 / (1.0 - u) * ratio
+        bb = np.maximum(bb, 0.0)
+        return bb
+
+    @staticmethod
+    def _compute_a(
+        u: np.ndarray,
+        aw_0: float,
+        bbw_0: float,
+        wavelength: np.ndarray,
+        lambda_0: int
+    ) -> np.ndarray:
+        """
+        推导全波段总吸收系数 a(λ)。
+
+        由 u = b_b / (a + b_b) 推导：
+            a = b_b * (1 - u) / u
+
+        Parameters
+        ----------
+        u : np.ndarray
+            中间变量 u。
+        aw_0 : float
+            λ₀ 处的纯水吸收系数。
+        bbw_0 : float
+            λ₀ 处的纯水后向散射系数。
+        wavelength : np.ndarray
+            全波段波长数组。
+        lambda_0 : int
+            参考波长（锚点）。
+
+        Returns
+        -------
+        np.ndarray
+            全波段总吸收系数 a。
+        """
+        S = 0.5 if lambda_0 < 600 else 0.0
+        wavelength = np.asarray(wavelength, dtype=np.float64)
+        ratio = (float(lambda_0) / wavelength) ** S
+        bbw = bbw_0 * ratio
+        with np.errstate(divide='ignore', invalid='ignore'):
+            a = bbw * (1.0 - u) / u + aw_0
+        a[~np.isfinite(a)] = np.nan
+        return a
+
+    # ------------------------------------------------------------------
+    # 主入口
+    # ------------------------------------------------------------------
+
+    def run_inversion(
+        self,
+        wavelengths: np.ndarray,
+        Rrs_spectrum: np.ndarray,
+        lambda_0: int
+    ) -> dict:
+        """
+        执行 QAA 核心反演。
+
+        Parameters
+        ----------
+        wavelengths : np.ndarray
+            光谱波长数组（nm），形状 (n_bands,) 或 (n_samples, n_bands)。
+        Rrs_spectrum : np.ndarray
+            水面遥感反射率光谱数据，形状 (n_bands,) 或 (n_samples, n_bands)。
+            若为 2D，每行为一个样本的光谱。
+        lambda_0 : int
+            参考波长（锚点），用于查表获取纯水 IOPs。
+
+        Returns
+        -------
+        dict
+            包含以下键的字典：
+              - wavelengths : 波长数组
+              - Rrs : 输入 Rrs
+              - r_rs_subsurface : 水下遥感反射率
+              - u : 中间变量
+              - a_lambda : 总吸收系数 a(λ)
+              - bb_lambda : 后向散射系数 b_b(λ)
+              - aw : λ₀ 处纯水吸收
+              - bbw : λ₀ 处纯水后向散射
+        """
+        wavelengths = np.asarray(wavelengths, dtype=np.float64)
+        Rrs_spectrum = np.asarray(Rrs_spectrum, dtype=np.float64)
+
+        if Rrs_spectrum.ndim == 1:
+            Rrs_spectrum = Rrs_spectrum[np.newaxis, :]
+
+        aw_0, bbw_0 = self._get_pure_water_iops(lambda_0)
+
+        results = []
+        for row in Rrs_spectrum:
+            rrs_sub = self._rrs_to_rrs_subsurface(row)
+            u = self._compute_u(rrs_sub)
+            bb = self._compute_bb(u, bbw_0, wavelengths, lambda_0)
+            a = self._compute_a(u, aw_0, bbw_0, wavelengths, lambda_0)
+            results.append({
+                'wavelengths': wavelengths,
+                'Rrs': row,
+                'r_rs_subsurface': rrs_sub,
+                'u': u,
+                'a_lambda': a,
+                'bb_lambda': bb,
+                'aw_0': aw_0,
+                'bbw_0': bbw_0,
+            })
+
+        if len(results) == 1:
+            return results[0]
+        return results
+
+    def invert_to_csv(
+        self,
+        wavelengths: np.ndarray,
+        Rrs_spectrum: np.ndarray,
+        lambda_0: int,
+        output_csv: str,
+        wavelength_col: str = "Wavelength",
+        sample_ids: Optional[list] = None
+    ) -> str:
+        """
+        执行反演并将结果保存为 CSV 文件。
+
+        Parameters
+        ----------
+        wavelengths : np.ndarray
+            波长数组（n_bands,）。
+        Rrs_spectrum : np.ndarray
+            光谱数据，形状 (n_bands,) 或 (n_samples, n_bands)。
+        lambda_0 : int
+            参考波长。
+        output_csv : str
+            输出 CSV 文件路径。
+        wavelength_col : str
+            输出 CSV 中波长列的列名前缀。
+        sample_ids : list, optional
+            样本 ID 列表（若为 None，使用 row_0, row_1, ...）。
+
+        Returns
+        -------
+        str
+            输出文件路径。
+        """
+        wavelengths = np.asarray(wavelengths, dtype=np.float64)
+        Rrs_spectrum = np.asarray(Rrs_spectrum, dtype=np.float64)
+
+        if Rrs_spectrum.ndim == 1:
+            Rrs_spectrum = Rrs_spectrum[np.newaxis, :]
+
+        n_samples = Rrs_spectrum.shape[0]
+        if sample_ids is None:
+            sample_ids = [f"sample_{i}" for i in range(n_samples)]
+
+        aw_0, bbw_0 = self._get_pure_water_iops(lambda_0)
+
+        rows_out = []
+        for i, row in enumerate(Rrs_spectrum):
+            rrs_sub = self._rrs_to_rrs_subsurface(row)
+            u = self._compute_u(rrs_sub)
+            bb = self._compute_bb(u, bbw_0, wavelengths, lambda_0)
+            a = self._compute_a(u, aw_0, bbw_0, wavelengths, lambda_0)
+            for j, wl in enumerate(wavelengths):
+                rows_out.append({
+                    'sample_id': sample_ids[i],
+                    'Wavelength': wl,
+                    'Rrs': row[j],
+                    'r_rs': rrs_sub[j],
+                    'u': u[j],
+                    'a_lambda': a[j],
+                    'bb_lambda': bb[j],
+                })
+
+        df = pd.DataFrame(rows_out)
+        os.makedirs(os.path.dirname(output_csv) or '.', exist_ok=True)
+        df.to_csv(output_csv, index=False, float_format='%.8f')
+        return output_csv

src/core/algorithms/waterindex_inversion.py

@@ -0,0 +1,22 @@
+# -*- coding: utf-8 -*-
+"""
+水色指数反演模块（包入口）
+
+从 waterindex.csv 读取公式，对去耀斑 BSQ 高光谱影像进行全图矩阵运算，
+输出带完整坐标信息的 GeoTIFF。
+
+公式格式（waterindex.csv）：
+  - 波长占位符：w{nm}，如 w686, w708, w665
+  - 支持混合大小写：w686 / W665 均可
+  - 示例：NDCI = (w708 - w665) / (w708 + w665)
+
+输出：
+  - GeoTIFF (Float32)，LZW 压缩，带 Tile
+  - 完整克隆原始 BSQ 的 GeoTransform / Projection / NoData
+  - Step 14 可直接用 rasterio 读取数组和空间范围
+"""
+
+# 重新导出 WaterIndexProcessor（向后兼容所有已有 import）
+from src.core.algorithms.waterindex_inversion import WaterIndexProcessor
+
+__all__ = ['WaterIndexProcessor']

src/core/algorithms/waterindex_inversion/__init__.py

@@ -0,0 +1,646 @@
+# -*- coding: utf-8 -*-
+"""
+水色指数反演模块
+
+直接读取去耀斑高光谱 BSQ 影像，应用 waterindex.csv 中的公式，
+输出各水质参数指数的 GeoTIFF 栅格图像。
+
+公式格式（waterindex.csv）：
+  - 波长占位符：w{nm}，如 w686, w708, w665
+  - 支持混合大小写：w686 / W665 均可
+  - 示例：NDCI = (w708 - w665) / (w708 + w665)
+          BGA_Am09KBBI = (w686 - w658) / (w686 + w658)
+
+输出：
+  - GeoTIFF (Float32)，LZW 压缩，带 Tile
+  - 完整克隆原始 BSQ 的 GeoTransform / Projection / NoData
+  - Step 14 可直接用 rasterio 读取进行克里金插值
+"""
+
+from __future__ import annotations
+
+import csv
+import os
+import re
+import sys
+import time
+import traceback
+from pathlib import Path
+from typing import Any, Callable, Dict, List, Optional, Tuple
+
+import numpy as np
+from osgeo import gdal, osr
+
+# GDAL 驱动注册
+gdal.UseExceptions()
+
+
+# ------------------------------------------------------------------
+# 公共工具
+# ------------------------------------------------------------------
+
+def _get_resource_path(relative_path: str) -> str:
+    """获取 waterindex.csv 等资源的绝对路径，兼容 PyInstaller 打包。"""
+    if hasattr(sys, '_MEIPASS'):
+        base = sys._MEIPASS
+    else:
+        base = os.path.abspath(
+            os.path.join(os.path.dirname(os.path.dirname(__file__)), '..', '..', '..')
+        )
+    return os.path.join(base, relative_path)
+
+
+# ------------------------------------------------------------------
+# WaterIndexProcessor
+# ------------------------------------------------------------------
+
+class WaterIndexProcessor:
+    """
+    水色指数处理器
+
+    读取 waterindex.csv 中的公式，应用于 BSQ 高光谱影像，
+    输出带完整坐标信息的 GeoTIFF 指数图。
+
+    核心能力：
+      - 公式解析：w{nm} 占位符 → 实际波段 2D numpy 数组
+      - 矩阵运算：全影像批量计算，无需逐点循环
+      - 地理信息保持：克隆原始 BSQ 的 GeoTransform / Projection
+      - NoData 处理：运算中产生的 NaN/Inf 统一标记为 -9999
+    """
+
+    # 内置安全命名空间（公式 eval 白名单）
+    _SAFE_NS: Dict[str, Any] = {
+        'np': np,
+        'nan': np.nan,
+        'inf': np.inf,
+        'pi': np.pi,
+        'e': np.e,
+    }
+
+    def __init__(self, waterindex_csv_path: Optional[str] = None):
+        """
+        Parameters
+        ----------
+        waterindex_csv_path : str, optional
+            waterindex.csv 路径。
+            若为 None，尝试从默认位置加载：
+              1. src/gui/model/waterindex.csv（开发环境）
+              2. _MEIPASS/src/gui/model/waterindex.csv（打包环境）
+        """
+        self.csv_path: Optional[str] = None
+        self.formulas: List[Dict[str, Any]] = []
+
+        if waterindex_csv_path:
+            self.csv_path = waterindex_csv_path
+        else:
+            candidates = [
+                os.path.join(os.path.dirname(__file__), '..', '..', 'gui', 'model', 'waterindex.csv'),
+                os.path.join(os.path.dirname(__file__), '..', '..', '..', 'gui', 'model', 'waterindex.csv'),
+            ]
+            for p in candidates:
+                if os.path.isfile(p):
+                    self.csv_path = p
+                    break
+
+        if self.csv_path:
+            self._parse_csv()
+        else:
+            self.formulas = []
+
+    # ------------------------------------------------------------------
+    # 公式加载
+    # ------------------------------------------------------------------
+
+    def _parse_csv(self) -> None:
+        """解析 waterindex.csv，加载所有公式。"""
+        if not os.path.isfile(self.csv_path):
+            raise FileNotFoundError(f"公式配置文件不存在: {self.csv_path}")
+
+        # ★★★ 防止多次调用时公式翻倍叠加 ★★★
+        self.formulas.clear()
+
+        with open(self.csv_path, 'r', encoding='utf-8-sig') as f:
+            reader = csv.DictReader(f)
+            for row in reader:
+                self.formulas.append(dict(row))
+
+        print(f"[WaterIndexProcessor] 加载 {len(self.formulas)} 条公式 ← {self.csv_path}")
+
+    def reload(self, waterindex_csv_path: str) -> None:
+        """重新加载公式配置文件。"""
+        self.csv_path = waterindex_csv_path
+        self._parse_csv()
+
+    # ------------------------------------------------------------------
+    # 公式查询
+    # ------------------------------------------------------------------
+
+    def list_formulas(self) -> List[Dict[str, Any]]:
+        """返回所有公式的列表。"""
+        return list(self.formulas)
+
+    def list_formula_names(self) -> List[str]:
+        """返回所有公式名称列表。"""
+        return [f.get('Formula_Name', '') for f in self.formulas]
+
+    def get_formula(self, name: str) -> Optional[Dict[str, Any]]:
+        """按名称查找单个公式。"""
+        for f in self.formulas:
+            if f.get('Formula_Name', '').strip() == name.strip():
+                return f
+        return None
+
+    def list_categories(self) -> List[str]:
+        """返回所有公式类别（去重排序）。"""
+        cats = set()
+        for f in self.formulas:
+            c = f.get('Category', '').strip()
+            if c:
+                cats.add(c)
+        return sorted(cats)
+
+    def get_formulas_by_category(self, category: str) -> List[Dict[str, Any]]:
+        """按类别筛选公式。"""
+        return [f for f in self.formulas
+                if f.get('Category', '').strip().lower() == category.strip().lower()]
+
+    # ------------------------------------------------------------------
+    # 影像元数据
+    # ------------------------------------------------------------------
+
+    def get_image_metadata(self, bsq_path: str, hdr_path: Optional[str] = None) -> Dict[str, Any]:
+        """获取影像元数据（GDAL + ENVI HDR 双重保障）。
+
+        Parameters
+        ----------
+        bsq_path : str
+            BSQ 影像路径
+        hdr_path : str, optional
+            ENVI HDR 路径（None → 自动构造）
+
+        Returns
+        -------
+        dict
+            含 keys: width, height, bands, wavelengths, wavelength_range,
+                     geotransform, projection, driver
+        """
+        meta: Dict[str, Any] = {}
+
+        # 1. GDAL 优先（获取空间信息）
+        try:
+            ds = gdal.Open(bsq_path, gdal.GA_ReadOnly)
+            if ds is not None:
+                meta['width'] = ds.RasterXSize
+                meta['height'] = ds.RasterYSize
+                meta['bands'] = ds.RasterCount
+                meta['driver'] = ds.GetDriver().ShortName
+                gt = ds.GetGeoTransform()
+                proj = ds.GetProjection()
+                if gt and gt != (0, 1, 0, 0, 0, 1):
+                    meta['geotransform'] = gt
+                if proj:
+                    meta['projection'] = proj
+                ds = None
+        except Exception:
+            pass
+
+        # 2. HDR 补充波长信息
+        if hdr_path is None:
+            hdr_path = os.path.splitext(bsq_path)[0] + '.hdr'
+            if not os.path.isfile(hdr_path):
+                hdr_path_alt = os.path.splitext(bsq_path)[0] + '.HDR'
+                if os.path.isfile(hdr_path_alt):
+                    hdr_path = hdr_path_alt
+
+        if os.path.isfile(hdr_path):
+            wl = self._parse_wavelengths_from_hdr(hdr_path)
+            if wl:
+                meta['wavelengths'] = wl
+                if len(wl) >= 2:
+                    meta['wavelength_range'] = f"{wl[0]:.1f}–{wl[-1]:.1f} nm ({len(wl)} 波段)"
+                elif meta.get('bands', 0) > 0:
+                    meta['wavelength_range'] = f"{meta['bands']} 波段（波长信息缺失）"
+
+        return meta
+
+    @staticmethod
+    def _parse_wavelengths_from_hdr(hdr_path: str) -> Optional[List[float]]:
+        """从 ENVI .hdr 文件中解析波长列表。"""
+        try:
+            with open(hdr_path, 'r', encoding='utf-8', errors='ignore') as f:
+                content = f.read()
+
+            # 格式1：wavelength = { 400, 401, ... }
+            m = re.search(r'wavelength\s*=\s*\{([^}]+)\}', content, re.DOTALL)
+            if m:
+                vals = [float(v) for v in re.findall(r'[\d.]+', m.group(1)) if v.strip()]
+                if vals:
+                    return vals
+
+            # 格式2：逐行罗列
+            wavelengths: List[float] = []
+            in_wl = False
+            for line in content.split('\n'):
+                line = line.strip()
+                if line.startswith('wavelength'):
+                    in_wl = True
+                    continue
+                if in_wl:
+                    if line.startswith('{'):
+                        continue
+                    try:
+                        wavelengths.append(float(line))
+                    except ValueError:
+                        if '}' in line:
+                            in_wl = False
+            return wavelengths if wavelengths else None
+        except Exception:
+            return None
+
+    # ------------------------------------------------------------------
+    # 公式解析：w{nm} 占位符 → 实际波段数据
+    # ------------------------------------------------------------------
+
+    def _find_nearest_band_index(self, target_wv: float,
+                                  wavelengths: List[float]) -> int:
+        """找到最接近目标波长的 GDAL 波段索引（1-based）。"""
+        if not wavelengths:
+            raise ValueError("波长列表为空，无法匹配波段")
+        nearest = min(range(len(wavelengths)),
+                      key=lambda i: abs(wavelengths[i] - target_wv))
+        return nearest + 1  # GDAL 波段从 1 开始
+
+    def _parse_formula_wavelengths(self, formula: str) -> List[int]:
+        """从公式字符串中提取所有波长值（去重，int）。"""
+        raw = re.findall(r'[wW](\d+)', formula)
+        seen = set()
+        result: List[int] = []
+        for r in raw:
+            v = int(r)
+            if v not in seen:
+                seen.add(v)
+                result.append(v)
+        return result
+
+    def _eval_formula_fast(self, formula: str,
+                           band_data: Dict[int, np.ndarray]) -> Optional[np.ndarray]:
+        """快速公式求值（预处理后直接 eval）。
+
+        band_data: {波长int: 2D 数组}
+        formula 示例: "(w708 - w665) / (w708 + w665)"
+        """
+        # 预处理：w708 → _B708（避免与 Python 关键字冲突）
+        processed = re.sub(r'[wW](\d+)', r'_B\1', formula)
+
+        # 构建局部变量表：_B708 = band_data[708]
+        local_vars = {f"_B{wv}": arr for wv, arr in band_data.items()}
+        local_vars.update(self._SAFE_NS)
+
+        try:
+            result = eval(processed, {"__builtins__": {}}, local_vars)
+            return result
+        except Exception as e:
+            print(f"  ⚠ 公式求值失败 [{formula}]: {e}")
+            return None
+
+    # ------------------------------------------------------------------
+    # 单波段读取（带 NoData 处理）
+    # ------------------------------------------------------------------
+
+    @staticmethod
+    def _read_band_as_float(bsq_path: str, band_idx: int) -> np.ndarray:
+        """读取 BSQ 指定波段（1-based），返回 float64，NaN 替换 NoData。"""
+        ds = gdal.Open(bsq_path, gdal.GA_ReadOnly)
+        if ds is None:
+            raise RuntimeError(f"无法用 GDAL 打开影像: {bsq_path}")
+
+        band = ds.GetRasterBand(band_idx)
+        arr = band.ReadAsArray()
+        nodata = band.GetNoDataValue()
+        ds = None
+
+        arr = arr.astype(np.float64)
+        if nodata is not None:
+            arr = np.where(arr == nodata, np.nan, arr)
+
+        return arr
+
+    # ------------------------------------------------------------------
+    # 核心处理：逐公式矩阵运算 + GeoTIFF 输出
+    # ------------------------------------------------------------------
+
+    def process_bsq(
+        self,
+        bsq_path: str,
+        hdr_path: Optional[str] = None,
+        output_dir: Optional[str] = None,
+        formula_names: Optional[List[str]] = None,
+        water_mask: Optional[np.ndarray] = None,
+        nodata_value: float = -9999.0,
+        progress_callback: Optional[Callable[[str, float], None]] = None,
+    ) -> Dict[str, str]:
+        """逐公式处理 BSQ 影像，输出 GeoTIFF。
+
+        Parameters
+        ----------
+        bsq_path : str
+            去耀斑 BSQ 影像路径
+        hdr_path : str, optional
+            ENVI HDR 文件路径（None → 自动构造）
+        output_dir : str, optional
+            输出目录（None → 与 bsq_path 同目录下的 10_WaterIndex_Images/）
+        formula_names : list, optional
+            要处理的公式名列表（None → 处理全部）
+        water_mask : np.ndarray, optional
+            水域掩膜数组（与 BSQ 同形状），掩膜值为 0 表示陆地，
+            将被强制赋值为 nodata_value
+        nodata_value : float
+            NoData 标记值
+        progress_callback : callable, optional
+            回调 (msg: str, pct: float)
+
+        Returns
+        -------
+        dict
+            {公式名: 输出 GeoTIFF 路径}
+        """
+        # ── 自动构造 HDR 路径 ────────────────────────────────────────────
+        if hdr_path is None:
+            hdr_path = os.path.splitext(bsq_path)[0] + '.hdr'
+            if not os.path.isfile(hdr_path):
+                hdr_path_alt = os.path.splitext(bsq_path)[0] + '.HDR'
+                if os.path.isfile(hdr_path_alt):
+                    hdr_path = hdr_path_alt
+
+        # ── 自动构造输出目录 ────────────────────────────────────────────
+        if output_dir is None:
+            output_dir = os.path.join(os.path.dirname(bsq_path), '10_WaterIndex_Images')
+        os.makedirs(output_dir, exist_ok=True)
+
+        def progress(msg: str, pct: float):
+            if progress_callback:
+                progress_callback(msg, pct)
+
+        # ── 获取影像元数据 ───────────────────────────────────────────────
+        progress("正在打开影像并读取元数据…", 2)
+        meta = self.get_image_metadata(bsq_path, hdr_path)
+
+        width = meta.get('width', 0)
+        height = meta.get('height', 0)
+        n_bands = meta.get('bands', 0)
+        wavelengths = meta.get('wavelengths', [])
+        geotransform = meta.get('geotransform')
+        projection = meta.get('projection')
+
+        if n_bands == 0 or width == 0 or height == 0:
+            raise ValueError(f"影像元数据无效，无法处理: {bsq_path}")
+
+        if not wavelengths:
+            raise ValueError(f"无法从 {hdr_path} 读取波长信息，公式无法解析")
+
+        progress(
+            f"影像: {width}×{height}像素, {n_bands}波段, "
+            f"波长 {wavelengths[0]:.1f}–{wavelengths[-1]:.1f}nm",
+            5
+        )
+
+        # ── 过滤要处理的公式 ──────────────────────────────────────────────
+        if formula_names:
+            formulas_to_run = [
+                f for f in self.formulas
+                if f.get('Formula_Name', '').strip() in formula_names
+            ]
+        else:
+            formulas_to_run = list(self.formulas)
+
+        results: Dict[str, str] = {}
+        total = len(formulas_to_run)
+
+        # ── 逐公式处理 ───────────────────────────────────────────────────
+        for i, formula_row in enumerate(formulas_to_run):
+            fname = formula_row.get('Formula_Name', '').strip()
+            fstr = formula_row.get('Formula', '').strip()
+            category = formula_row.get('Category', '').strip()
+            ftype = formula_row.get('Formula_Type', '').strip()
+
+            if not fname or not fstr:
+                continue
+
+            progress(
+                f"[{i + 1}/{total}] {fname} ({category})",
+                5 + 90 * i / total
+            )
+
+            try:
+                # 1) 提取公式所需的波长列表
+                required_wvs = self._parse_formula_wavelengths(fstr)
+
+                # 2) 按需读取波段数据（相同波长只读一次）
+                band_data: Dict[int, np.ndarray] = {}
+                for wv in required_wvs:
+                    if wv not in band_data:
+                        band_idx = self._find_nearest_band_index(wv, wavelengths)
+                        if not (0 < band_idx <= n_bands):
+                            print(f"  ⚠ 公式 '{fname}' 引用波段 {band_idx}，超出范围 ({n_bands})，跳过")
+                            raise ValueError(f"波段 {band_idx} 超出影像范围")
+                        band_data[wv] = self._read_band_as_float(bsq_path, band_idx)
+
+                # 3) 矩阵运算
+                index_arr = self._eval_formula_fast(fstr, band_data)
+                if index_arr is None:
+                    print(f"  ⚠ 公式 '{fname}' 计算失败，跳过")
+                    continue
+
+                # 4) NoData 处理：NaN / Inf → nodata_value
+                index_arr = np.where(np.isfinite(index_arr), index_arr, nodata_value)
+
+                # 4b) 水域掩膜拦截：陆地像素（mask==0）强制赋 NoData
+                if water_mask is not None:
+                    land_pixels = (water_mask == 0)
+                    land_count = int(land_pixels.sum())
+                    if land_count > 0:
+                        index_arr = np.where(land_pixels, nodata_value, index_arr)
+                        print(f"  🗺 掩膜处理：陆地像素 {land_count:,} 个已设为 NoData")
+
+                # 5) 输出 GeoTIFF
+                safe_fname = re.sub(r'[^\w\u4e00-\u9fff-]', '_', fname)
+                out_tif = os.path.join(output_dir, f"{safe_fname}.tif")
+
+                self._write_geotiff(
+                    out_path=out_tif,
+                    data=index_arr,
+                    reference_bsq=bsq_path,
+                    nodata_value=nodata_value,
+                    description=f"{fname}|{category}|{ftype}|{fstr}",
+                )
+
+                results[fname] = out_tif
+                valid = index_arr[index_arr != nodata_value]
+                mean_val = float(np.mean(valid)) if valid.size else np.nan
+                print(f"  ✅ {fname} → {out_tif}  (mean={mean_val:.4f})")
+
+            except ValueError as ve:
+                print(f"  ⏭ 跳过 '{fname}': {ve}")
+                continue
+            except Exception as e:
+                print(f"  ❌ 公式 '{fname}' 失败: {e}\n{traceback.format_exc()}")
+                continue
+
+        progress(f"完成！共输出 {len(results)} / {total} 个指数图", 100)
+        return results
+
+    def _write_geotiff(
+        self,
+        out_path: str,
+        data: np.ndarray,
+        reference_bsq: str,
+        nodata_value: float = -9999.0,
+        description: str = "",
+    ) -> None:
+        """将数组写入 GeoTIFF，克隆原始 BSQ 的地理信息。
+
+        Parameters
+        ----------
+        out_path : str
+            输出 GeoTIFF 路径
+        data : np.ndarray
+            2D 数据数组（height, width）
+        reference_bsq : str
+            参考 BSQ 影像路径（用于克隆 GeoTransform / Projection）
+        nodata_value : float
+            NoData 标记值
+        description : str
+            GDAL 数据集描述
+        """
+        height, width = data.shape
+
+        driver = gdal.GetDriverByName('GTiff')
+        if driver is None:
+            raise RuntimeError("GDAL GTiff 驱动不可用")
+
+        out_ds = driver.Create(
+            out_path,
+            width, height,
+            1,
+            gdal.GDT_Float32,
+            options=['COMPRESS=LZW', 'TILED=YES', 'BIGTIFF=IF_SAFER'],
+        )
+        if out_ds is None:
+            raise RuntimeError(f"无法创建 GeoTIFF: {out_path}")
+
+        # 写入数据
+        out_band = out_ds.GetRasterBand(1)
+        out_band.SetNoDataValue(nodata_value)
+        out_band.WriteArray(data)
+        out_band.FlushCache()
+
+        # 写入描述
+        if description:
+            out_band.SetDescription(description)
+
+        # ★★★ 克隆原始 BSQ 的 GeoTransform 和 Projection ★★★
+        ref_ds = gdal.Open(reference_bsq, gdal.GA_ReadOnly)
+        if ref_ds is not None:
+            gt = ref_ds.GetGeoTransform()
+            proj = ref_ds.GetProjection()
+            if gt and gt != (0, 1, 0, 0, 0, 1):
+                out_ds.SetGeoTransform(gt)
+            if proj:
+                out_ds.SetProjection(proj)
+            ref_ds = None
+
+        out_ds = None
+
+    # ------------------------------------------------------------------
+    # Pipeline 入口（供 PipelineRunner 调用）
+    # ------------------------------------------------------------------
+
+    def run_inversion(
+        self,
+        deglint_img_path: str,
+        work_dir: str,
+        formula_csv_path: Optional[str] = None,
+        selected_formulas: Optional[List[str]] = None,
+        water_mask_path: Optional[str] = None,
+        nodata_value: float = -9999.0,
+        callback: Optional[Callable] = None,
+        **kwargs,
+    ) -> Dict[str, str]:
+        """Pipeline 入口方法。
+
+        Parameters
+        ----------
+        deglint_img_path : str
+            去耀斑影像 BSQ 路径
+        work_dir : str
+            工作目录
+        formula_csv_path : str, optional
+            waterindex.csv 路径（None → 使用初始化时的路径）
+        selected_formulas : list, optional
+            要处理的公式列表
+        water_mask_path : str, optional
+            水域掩膜路径（如 1_water_mask/water_mask.dat），
+            掩膜中为 0 的像素视为陆地区域，其指数值将被强制设为 NoData。
+        nodata_value : float
+            NoData 标记值，默认 -9999.0
+        callback : callable, optional
+            进度回调
+
+        Returns
+        -------
+        dict
+            {公式名: 输出 GeoTIFF 路径}
+        """
+        # 重新加载公式（如指定了新路径）
+        if formula_csv_path:
+            self.reload(formula_csv_path)
+        elif not self.formulas:
+            raise RuntimeError("WaterIndexProcessor 未加载公式，请指定 formula_csv_path")
+
+        def notify(msg: str, pct: float):
+            if callback:
+                callback(msg, pct)
+
+        notify("开始水色指数反演", 0)
+
+        bsq_path = deglint_img_path
+        hdr_path = os.path.splitext(bsq_path)[0] + '.hdr'
+        if not os.path.isfile(hdr_path):
+            hdr_path_alt = os.path.splitext(bsq_path)[0] + '.HDR'
+            if os.path.isfile(hdr_path_alt):
+                hdr_path = hdr_path_alt
+
+        output_dir = os.path.join(work_dir, "10_WaterIndex_Images")
+
+        # ── 加载水域掩膜（可选）───────────────────────────────────────
+        water_mask: Optional[np.ndarray] = None
+        if water_mask_path:
+            if os.path.isfile(water_mask_path):
+                try:
+                    import rasterio
+                    with rasterio.open(water_mask_path) as msrc:
+                        water_mask = msrc.read(1)
+                    print(f"[run_inversion] 水域掩膜已加载: {water_mask_path}，"
+                          f"形状={water_mask.shape}，"
+                          f"陆地区域(0)={int((water_mask == 0).sum())}，"
+                          f"水区域(>0)={int((water_mask > 0).sum())}")
+                except Exception as mask_err:
+                    print(f"[run_inversion] ⚠ 掩膜加载失败，跳过掩膜处理: {mask_err}")
+                    water_mask = None
+            else:
+                print(f"[run_inversion] ⚠ 水域掩膜文件不存在: {water_mask_path}，跳过掩膜处理")
+
+        notify("水色指数处理中…", 20)
+
+        results = self.process_bsq(
+            bsq_path=bsq_path,
+            hdr_path=hdr_path,
+            output_dir=output_dir,
+            formula_names=selected_formulas,
+            water_mask=water_mask,
+            nodata_value=nodata_value,
+            progress_callback=lambda m, p: notify(m, 20 + 70 * p / 100),
+        )
+
+        notify("水色指数反演完成", 100)
+        return results

src/core/glint_removal/get_spectral-test.py

@@ -899,11 +899,11 @@ def get_spectral_in_coor(imgpath, coorpath, outpath, radius=0, flare_path=None,
 if __name__ == '__main__':
     # 在这里直接设置参数
     imgpath = r"D:\BaiduNetdiskDownload\yaobao\result3.bsq"# BIL格式影像文件路径
-    coorpath = r"E:\code\WQ\封装\work_dir\4_processed_data\processed_data.csv"# CSV格式坐标文件路径（第1、2列为纬度和经度）
+    coorpath = r"E:\code\WQ\封装\work_dir\5_Data_Cleaning\processed_data.csv"# CSV格式坐标文件路径（第1、2列为纬度和经度）
     output_path = r"E:\code\WQ\封装\test/yangdian_output.csv"  # CSV格式输出文件路径
     
     radius = 5  # 采样半径（像素），0表示单点采样，>0表示半径内平均
-    flare_path = r"E:\code\WQ\封装\work_dir\2_glint\severe_glint_area.dat"  # 耀斑掩膜文件路径（可选，None表示不使用）
+    flare_path = r"E:\code\WQ\封装\work_dir\2_Glint_Detection\severe_glint_area.dat"  # 耀斑掩膜文件路径（可选，None表示不使用）
     boundary_path ="D:\BaiduNetdiskDownload\yaobao\water_mask.dat"  # 边界掩膜文件路径（可选，None表示不使用）
     source_epsg = 4326  # 源坐标系EPSG代码，默认为4326 (WGS84地理坐标系)
     

src/core/glint_removal/get_spectral.py

@@ -2,6 +2,7 @@ from osgeo import gdal, osr
 import numpy as np
 import pandas as pd
 import os
+import re
 import spectral
 from math import sin, cos, tan, sqrt, radians
 
@@ -212,16 +213,83 @@ def load_mask_file(mask_path):
 def get_hdr_file_path(file_path):
     """
     获取HDR文件路径
-    
+
     Args:
         file_path: 影像文件路径
-        
+
     Returns:
         HDR文件路径
     """
     return os.path.splitext(file_path)[0] + ".hdr"
 
 
+def load_wavelength_columns(imgpath, num_bands):
+    """
+    加载 wavelength 列名（鲁棒版：三级回退）
+
+    优先级：
+      1) spectral.envi.read_envi_header（标准库解析，依赖 ENVI 头完整性）
+      2) 纯文本暴力解析 .hdr（兜底，绕过 spectral 对 band names / 波段数一致性的校验）
+         —— 解决 .hdr 中 band names 数量与 bands 不符导致的标准库解析失败问题
+      3) 最后回退：band_1, band_2, ..., band_N
+
+    Args:
+        imgpath: 影像文件路径（.bsq / .bil / .bip 等）
+        num_bands: 影像实际波段数（用于回退列名长度 & 不一致警告）
+
+    Returns:
+        spectral_columns: 长度为 num_bands 的字符串列表（与原代码列名格式一致：纯数字字符串）
+    """
+    hdr_path = get_hdr_file_path(imgpath)
+
+    # 1) 标准库解析
+    try:
+        in_hdr_dict = spectral.envi.read_envi_header(hdr_path)
+        wavelengths = np.array(in_hdr_dict['wavelength']).astype('float64')
+        spectral_columns = [str(wl) for wl in wavelengths]
+        print(f"[wavelength] 标准库解析成功，从 {hdr_path} 提取 {len(spectral_columns)} 个波长")
+        if len(spectral_columns) != num_bands:
+            print(f"[wavelength] 警告: 解析波长数 ({len(spectral_columns)}) 与影像波段数 ({num_bands}) 不一致，将以 num_bands 为准截断/补齐")
+        if len(spectral_columns) > num_bands:
+            spectral_columns = spectral_columns[:num_bands]
+        elif len(spectral_columns) < num_bands:
+            spectral_columns = spectral_columns + [f"band_{j+1}" for j in range(len(spectral_columns), num_bands)]
+        return spectral_columns
+    except Exception as e_std:
+        print(f"[wavelength] 标准库解析失败: {str(e_std)}，将尝试文本兜底解析")
+
+    # 2) 兜底：纯文本暴力解析
+    try:
+        if not os.path.isfile(hdr_path):
+            print(f"[wavelength] 文本兜底失败: {hdr_path} 不存在")
+        else:
+            with open(hdr_path, 'r', encoding='utf-8', errors='ignore') as f:
+                hdr_text = f.read()
+            pattern = r'wavelength\s*=\s*\{([^}]+)\}'
+            m = re.search(pattern, hdr_text, flags=re.IGNORECASE | re.DOTALL)
+            if m:
+                inner = m.group(1)
+                tokens = [t.strip() for t in inner.split(',') if t.strip()]
+                if tokens:
+                    if len(tokens) != num_bands:
+                        print(f"[wavelength] 文本解析波长数 ({len(tokens)}) 与影像波段数 ({num_bands}) 不一致，将以 num_bands 为准截断/补齐")
+                    if len(tokens) > num_bands:
+                        tokens = tokens[:num_bands]
+                    elif len(tokens) < num_bands:
+                        tokens = tokens + [f"band_{j+1}" for j in range(len(tokens), num_bands)]
+                    print(f"[wavelength] 文本暴力解析成功，从 {hdr_path} 提取 {len(tokens)} 个真实波长")
+                    return tokens
+                print(f"[wavelength] 文本兜底: 已匹配到 wavelength = {{ ... }}，但内部为空")
+            else:
+                print(f"[wavelength] 文本兜底: 未在 {hdr_path} 中匹配到 wavelength = {{ ... }} 字段")
+    except Exception as e_txt:
+        print(f"[wavelength] 文本兜底解析异常: {str(e_txt)}")
+
+    # 3) 全部失败，最后回退
+    print(f"[wavelength] 所有解析路径均失败，回退到 band_1..band_{num_bands}")
+    return ["band_" + str(j + 1) for j in range(num_bands)]
+
+
 def calculate_utm_zone(longitude):
     """
     根据经度计算UTM分区号
@@ -473,9 +541,56 @@ def get_spectral_in_coor(imgpath, coorpath, outpath, radius=0, flare_path=None,
     for i in range(min(3, coor_data.shape[0])):
         print(f"  行{i + 1}: {coor_data[i, :min(5, coor_data.shape[1])]}")  # 只显示前5列
     
-    # 提取原始坐标
-    lat_array = coor_data[:, 0]  # 第1列是纬度
-    lon_array = coor_data[:, 1]  # 第2列是经度
+    # 提取原始坐标（使用智能坐标列检测）
+    lon_patterns = [
+        r'^lon', r'^lng', r'^longitude', r'经度', r'^x$', r'^utm_x$', r'^pixel_x$'
+    ]
+    lat_patterns = [
+        r'^lat', r'^latitude', r'纬度', r'^y$', r'^utm_y$', r'^pixel_y$'
+    ]
+
+    x_col_name, y_col_name = None, None
+
+    if coor_df is not None and hasattr(coor_df, 'columns'):
+        for col in coor_df.columns:
+            col_str = str(col).lower().strip()
+            if x_col_name is None and any(re.search(p, col_str) for p in lon_patterns):
+                x_col_name = col
+            if y_col_name is None and any(re.search(p, col_str) for p in lat_patterns):
+                y_col_name = col
+
+    if x_col_name and y_col_name and x_col_name in coor_df.columns and y_col_name in coor_df.columns:
+        lon_array = coor_df[x_col_name].values
+        lat_array = coor_df[y_col_name].values
+        print(f"💡 坐标列名检测: X/经度=[{x_col_name}], Y/纬度=[{y_col_name}]")
+    else:
+        numeric_cols = coor_df.select_dtypes(include=[np.number]).columns.tolist() if coor_df is not None else []
+        if len(numeric_cols) >= 2:
+            col1, col2 = numeric_cols[0], numeric_cols[1]
+            mean1 = coor_df[col1].head(10).mean()
+            mean2 = coor_df[col2].head(10).mean()
+            if abs(mean1) <= 90 and abs(mean2) > 90:
+                y_col_name, x_col_name = col1, col2
+                lon_array = coor_df[x_col_name].values
+                lat_array = coor_df[y_col_name].values
+            elif abs(mean2) <= 90 and abs(mean1) > 90:
+                x_col_name, y_col_name = col1, col2
+                lon_array = coor_df[x_col_name].values
+                lat_array = coor_df[y_col_name].values
+            else:
+                if mean1 > mean2:
+                    x_col_name, y_col_name = col1, col2
+                else:
+                    x_col_name, y_col_name = col2, col1
+                lon_array = coor_df[x_col_name].values
+                lat_array = coor_df[y_col_name].values
+            print(f"💡 触发智能数值推断坐标列: X/经度=[{x_col_name}], Y/纬度=[{y_col_name}]")
+        else:
+            if coor_data is not None and coor_data.shape[1] >= 3:
+                lat_array = coor_data[:, 1]
+                lon_array = coor_data[:, 2]
+            else:
+                raise Exception("坐标文件格式错误：需要至少2列数据，且最好包含坐标列名（如lon/lat/经度/纬度）")
     
     print(f"\n=== 原始坐标信息 ===")
     print(f"原始坐标范围: 经度 {np.min(lon_array):.6f} ~ {np.max(lon_array):.6f}, 纬度 {np.min(lat_array):.6f} ~ {np.max(lat_array):.6f}")
@@ -711,17 +826,8 @@ def get_spectral_in_coor(imgpath, coorpath, outpath, radius=0, flare_path=None,
         else:
             original_columns = []
     
-    # 读取波长信息，用作光谱列名
-    wavelengths = None
-    try:
-        in_hdr_dict = spectral.envi.read_envi_header(get_hdr_file_path(imgpath))
-        wavelengths = np.array(in_hdr_dict['wavelength']).astype('float64')
-        # 将波长值转换为字符串作为列名
-        spectral_columns = [str(wl) for wl in wavelengths]
-        print(f"成功读取波长信息，共 {len(spectral_columns)} 个波段")
-    except Exception as e:
-        print(f"警告: 无法读取波长信息 ({str(e)})，使用默认列名 band_1, band_2, ...")
-        spectral_columns = ["band_" + str(j + 1) for j in range(num_bands)]
+    # 读取波长信息，用作光谱列名（三级回退：spectral 解析 → 文本暴力解析 → band_N 兜底）
+    spectral_columns = load_wavelength_columns(imgpath, num_bands)
     
     # 构建输出列名（不包含前两列坐标列和UTM列）
     all_columns = original_columns + spectral_columns
@@ -758,11 +864,11 @@ def get_spectral_in_coor(imgpath, coorpath, outpath, radius=0, flare_path=None,
 if __name__ == '__main__':
     # 在这里直接设置参数
     imgpath = r"E:\code\WQ\封装\work_dir\3_deglint\deglint_goodman.bsq" # BIL格式影像文件路径
-    coorpath = r"E:\code\WQ\封装\work_dir\4_processed_data\processed_data.csv"# CSV格式坐标文件路径（第1、2列为纬度和经度）
-    output_path = r"E:\code\WQ\封装\work_dir\5_training_spectra/yangdian_output.csv"  # CSV格式输出文件路径
+    coorpath = r"E:\code\WQ\封装\work_dir\5_Data_Cleaning\processed_data.csv"# CSV格式坐标文件路径（第1、2列为纬度和经度）
+    output_path = r"E:\code\WQ\封装\work_dir\6_Spectral_Feature_Extraction/yangdian_output.csv"  # CSV格式输出文件路径
     
     radius = 5  # 采样半径（像素），0表示单点采样，>0表示半径内平均
-    flare_path = r"E:\code\WQ\封装\work_dir\2_glint\severe_glint_area.dat"  # 耀斑掩膜文件路径（可选，None表示不使用）
+    flare_path = r"E:\code\WQ\封装\work_dir\2_Glint_Detection\severe_glint_area.dat"  # 耀斑掩膜文件路径（可选，None表示不使用）
     boundary_path = r"D:\BaiduNetdiskDownload\yaobao\water_mask.dat"  # 边界掩膜文件路径（可选，None表示不使用）
     source_epsg = 4326  # 源坐标系EPSG代码，默认为4326 (WGS84地理坐标系)
     

src/core/handlers/__init__.py

@@ -0,0 +1,46 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+步骤处理器包
+
+将 WaterQualityInversionPipeline 的 14 个巨型 step* 方法
+拆分为独立的 Handler 类，每个 Handler 实现 BaseStepHandler 接口。
+
+调度器（PipelineScheduler）仅维护执行上下文并根据 step_key
+从注册表查找对应 Handler 执行，自身不再包含任何算法逻辑。
+"""
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+from src.core.handlers.step1_water_mask import Step1WaterMaskHandler
+from src.core.handlers.step2_glint_detection import Step2GlintDetectionHandler
+from src.core.handlers.step3_glint_removal import Step3GlintRemovalHandler
+from src.core.handlers.step4_sampling import Step4SamplingHandler
+from src.core.handlers.step5_process_csv import Step5ProcessCsvHandler
+from src.core.handlers.step6_extract_spectra import Step6ExtractSpectraHandler
+from src.core.handlers.step7_calc_indices import Step7CalcIndicesHandler
+from src.core.handlers.step8_ml_train import Step8MlTrainHandler
+from src.core.handlers.step9_ml_predict import Step9MlPredictHandler
+from src.core.handlers.step10_qaa_inversion import Step10QaaInversionHandler
+from src.core.handlers.step11_concentration import Step11ConcentrationHandler
+from src.core.handlers.step12_kriging import Step12KrigingHandler
+from src.core.handlers.step13_visualization import Step13VisualizationHandler
+from src.core.handlers.step14_report import Step14ReportHandler
+
+__all__ = [
+    'BaseStepHandler',
+    'PipelineContext',
+    'Step1WaterMaskHandler',
+    'Step2GlintDetectionHandler',
+    'Step3GlintRemovalHandler',
+    'Step4SamplingHandler',
+    'Step5ProcessCsvHandler',
+    'Step6ExtractSpectraHandler',
+    'Step7CalcIndicesHandler',
+    'Step8MlTrainHandler',
+    'Step9MlPredictHandler',
+    'Step10QaaInversionHandler',
+    'Step11ConcentrationHandler',
+    'Step12KrigingHandler',
+    'Step13VisualizationHandler',
+    'Step14ReportHandler',
+]

src/core/handlers/base.py

@@ -0,0 +1,282 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Handler 基类与 Pipeline 执行上下文
+
+BaseStepHandler —— 所有步骤 Handler 的抽象基类，定义统一的 execute 接口。
+PipelineContext  —— 在 Handler 之间传递的共享状态容器（路径、计时、回调等）。
+
+设计原则：
+- Handler 只负责"执行一个步骤的算法逻辑"，不管理调度/依赖/跳过。
+- Context 是 Handler 之间唯一的共享状态通道。
+- 调度器（PipelineScheduler）负责遍历 config、查找 Handler、调用 execute。
+"""
+
+from __future__ import annotations
+
+import time
+from abc import ABC, abstractmethod
+from datetime import datetime
+from pathlib import Path
+from typing import Any, Callable, Dict, List, Optional
+
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+import warnings
+
+warnings.filterwarnings('ignore')
+
+
+class PipelineContext:
+    """管道执行上下文 —— Handler 之间共享状态的唯一载体。
+
+    包含：
+    - 工作目录及子目录
+    - 中间结果路径（water_mask_path, glint_mask_path, ...）
+    - 步骤计时记录
+    - 回调函数（用于 GUI 进度通知）
+    - 可视化/报告生成器实例
+    """
+
+    def __init__(self, work_dir: str = "./work_dir"):
+        self.work_dir = Path(work_dir)
+        self.work_dir.mkdir(parents=True, exist_ok=True)
+
+        # ── 子目录 ──
+        self.water_mask_dir = self.work_dir / "1_water_mask"
+        self.glint_dir = self.work_dir / "2_Glint_Detection"
+        self.deglint_dir = self.work_dir / "3_deglint"
+        self.processed_data_dir = self.work_dir / "5_Data_Cleaning"
+        self.training_spectra_dir = self.work_dir / "6_Spectral_Feature_Extraction"
+        self.indices_dir = self.work_dir / "7_Water_Quality_Indices"
+        self.models_dir = self.work_dir / "8_Supervised_Model_Training"
+        self.non_empirical_models_dir = self.work_dir / "8_Non_Empirical_Regression"
+        self.custom_regression_dir = self.work_dir / "13_Custom_Regression"
+        self.sampling_dir = self.work_dir / "4_sampling"
+        self.prediction_dir = self.work_dir / "11_12_13_predictions"
+        self.visualization_dir = self.work_dir / "14_visualization"
+        self.reports_dir = self.work_dir / "reports"
+
+        for d in [self.water_mask_dir, self.glint_dir, self.deglint_dir,
+                  self.processed_data_dir, self.training_spectra_dir,
+                  self.indices_dir, self.models_dir, self.non_empirical_models_dir,
+                  self.custom_regression_dir, self.sampling_dir, self.prediction_dir,
+                  self.visualization_dir, self.reports_dir]:
+            d.mkdir(parents=True, exist_ok=True)
+
+        # ── 中间结果路径 ──
+        self.water_mask_path: Optional[str] = None
+        self.glint_mask_path: Optional[str] = None
+        self.interpolated_img_path: Optional[str] = None
+        self.deglint_img_path: Optional[str] = None
+        self.processed_csv_path: Optional[str] = None
+        self.training_csv_path: Optional[str] = None
+        self.indices_path: Optional[str] = None
+        self.custom_regression_path: Optional[str] = None
+        self.sampling_csv_path: Optional[str] = None
+        self.prediction_files: Dict[str, str] = {}
+        self.distribution_map_path: Optional[str] = None
+        self.qaa_output_path: Optional[str] = None
+        self.concentration_output_path: Optional[str] = None
+
+        # ── 计时 ──
+        self.step_timings: Dict[str, dict] = {}
+        self.pipeline_start_time: Optional[float] = None
+        self.pipeline_end_time: Optional[float] = None
+
+        # ── 回调 ──
+        self._callback: Optional[Callable] = None
+
+        # ── 可视化组件（延迟导入避免循环依赖）──
+        self._visualizer = None
+        self._report_generator = None
+        self._scatter_batch = None
+
+        # ── matplotlib 中文字体 ──
+        plt.rcParams['font.sans-serif'] = ['SimHei', 'Microsoft YaHei',
+                                            'DejaVu Sans', 'Arial Unicode MS']
+        plt.rcParams['axes.unicode_minus'] = False
+
+    # ═══════════════════════════════════════════════════════════
+    # 回调
+    # ═══════════════════════════════════════════════════════════
+
+    def set_callback(self, callback: Callable):
+        """设置回调函数，用于向 GUI 报告进度。
+
+        Args:
+            callback: 签名为 callback(step_name, status, message="")
+                      status: 'start' | 'completed' | 'skipped' | 'error' | 'info' | 'warning'
+        """
+        self._callback = callback
+
+    def notify(self, step_name: str, status: str, message: str = ""):
+        """通知回调函数。"""
+        if self._callback:
+            try:
+                self._callback(step_name, status, message)
+            except Exception as e:
+                print(f"回调函数执行失败: {e}")
+
+    # ═══════════════════════════════════════════════════════════
+    # 计时
+    # ═══════════════════════════════════════════════════════════
+
+    def record_step_time(self, step_name: str, start_time: float, end_time: float,
+                         status: str = "completed", error: Optional[str] = None):
+        elapsed = end_time - start_time
+        self.step_timings[step_name] = {
+            'start_time': datetime.fromtimestamp(start_time).strftime('%Y-%m-%d %H:%M:%S'),
+            'end_time': datetime.fromtimestamp(end_time).strftime('%Y-%m-%d %H:%M:%S'),
+            'elapsed_seconds': elapsed,
+            'elapsed_formatted': self._format_time(elapsed),
+            'status': status,
+            'error': error,
+        }
+
+    @staticmethod
+    def _format_time(seconds: float) -> str:
+        if seconds < 60:
+            return f"{seconds:.2f}秒"
+        elif seconds < 3600:
+            minutes = int(seconds // 60)
+            secs = seconds % 60
+            return f"{minutes}分{secs:.2f}秒"
+        else:
+            hours = int(seconds // 3600)
+            minutes = int((seconds % 3600) // 60)
+            secs = seconds % 60
+            return f"{hours}小时{minutes}分{secs:.2f}秒"
+
+    # ═══════════════════════════════════════════════════════════
+    # 可视化组件（延迟导入）
+    # ═══════════════════════════════════════════════════════════
+
+    @property
+    def visualizer(self):
+        if self._visualizer is None:
+            from src.postprocessing.visualization_reports import WaterQualityVisualization
+            self._visualizer = WaterQualityVisualization(str(self.visualization_dir))
+        return self._visualizer
+
+    @property
+    def report_generator(self):
+        if self._report_generator is None:
+            from src.postprocessing.visualization_reports import ReportGenerator
+            self._report_generator = ReportGenerator(str(self.reports_dir))
+        return self._report_generator
+
+    @property
+    def scatter_batch(self):
+        if self._scatter_batch is None:
+            from src.core.prediction.sctter_batch import WaterQualityScatterBatch
+            self._scatter_batch = WaterQualityScatterBatch()
+        return self._scatter_batch
+
+    # ═══════════════════════════════════════════════════════════
+    # 步骤输出目录查找（兼容旧接口）
+    # ═══════════════════════════════════════════════════════════
+
+    _STEP_OUTPUT_DIR_MAP: Optional[Dict[str, Path]] = None
+
+    def _ensure_step_dir_map(self) -> Dict[str, Path]:
+        if PipelineContext._STEP_OUTPUT_DIR_MAP is not None:
+            return PipelineContext._STEP_OUTPUT_DIR_MAP
+        wp = self.work_dir
+        m = {
+            'step1': wp / '1_water_mask',
+            'step2': wp / '2_Glint_Detection',
+            'step3': wp / '3_deglint',
+            'step4_sampling': wp / '4_sampling',
+            'step5_clean': wp / '5_Data_Cleaning',
+            'step6_feature': wp / '6_Spectral_Feature_Extraction',
+            'step7_index': wp / '7_Water_Quality_Indices',
+            'step8_ml_train': wp / '8_Supervised_Model_Training',
+            'step9_ml_predict': wp / '8_Non_Empirical_Regression',
+            'step10_watercolor': wp / '10_WaterIndex_Images',
+            'step11_map': wp / '14_visualization',
+            'step12_viz': wp / '14_visualization',
+            'step13_report': wp / '14_visualization',
+            'step11_predictions': wp / '11_12_13_predictions',
+            'step12_predictions': wp / '11_12_13_predictions',
+            'step13_predictions': wp / '11_12_13_predictions',
+            'custom_regression': wp / '13_Custom_Regression',
+            'prediction_dir': wp / '11_12_13_predictions',
+            'visualization': wp / '14_visualization',
+            'reports': wp / 'reports',
+            'step8': wp / '8_Supervised_Model_Training',
+            'step9': wp / '8_Non_Empirical_Regression',
+            'step10': wp / '10_WaterIndex_Images',
+            'step11': wp / '11_12_13_predictions',
+            'step12': wp / '13_Custom_Regression',
+            'step13': wp / 'reports',
+            'step14': wp / '14_visualization',
+        }
+        PipelineContext._STEP_OUTPUT_DIR_MAP = m
+        return m
+
+    def get_step_output_dir(self, step_name: str) -> Path:
+        mapping = self._ensure_step_dir_map()
+        key = (step_name or '').strip()
+        if key in mapping:
+            return mapping[key]
+        print(f"[PipelineContext.get_step_output_dir] 未知 step_name={key!r}，回退到 work_dir")
+        return self.work_dir
+
+
+class BaseStepHandler(ABC):
+    """步骤处理器抽象基类。
+
+    所有步骤 Handler 必须实现:
+    - step_key: 类属性，对应 config 中的 key（如 'step1', 'step2', ...）
+    - execute(context, config): 执行步骤逻辑，返回结果字典
+
+    用法示例::
+
+        class Step1WaterMaskHandler(BaseStepHandler):
+            step_key = 'step1'
+
+            def execute(self, ctx, config):
+                result = WaterMaskStep.run(...)
+                ctx.water_mask_path = result
+                return {'water_mask_path': result}
+    """
+
+    # 子类必须定义：对应 config 字典中的 key
+    step_key: str = None
+
+    @abstractmethod
+    def execute(self, context: PipelineContext, config: dict) -> dict:
+        """执行步骤逻辑。
+
+        Args:
+            context: 管道执行上下文（共享状态）
+            config:  该步骤的配置字典（即 config[self.step_key]）
+
+        Returns:
+            结果字典，包含该步骤产生的输出路径等信息。
+            调度器会将返回值合并到全局结果中。
+
+        Raises:
+            Exception: 任何异常都会由调度器捕获并记录。
+        """
+        ...
+
+    def _resolve_path(self, explicit: Optional[str], fallback: Optional[str],
+                      label: str = "path") -> Optional[str]:
+        """解析路径：优先使用显式传入值，否则回退到上下文中的缓存值。
+
+        Args:
+            explicit: 调用方显式传入的路径
+            fallback: 上下文中的缓存路径
+            label: 用于日志的标签
+
+        Returns:
+            解析后的路径，若两者均为 None 则返回 None
+        """
+        if explicit is not None:
+            return explicit
+        if fallback is not None:
+            return fallback
+        return None

src/core/handlers/pipeline_scheduler.py

@@ -0,0 +1,199 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+极简管道调度器
+
+替代原 WaterQualityInversionPipeline（2598 行上帝类）的调度核心。
+调度器自身不包含任何算法逻辑，仅负责：
+1. 维护 PipelineContext（共享状态）
+2. 根据 config key 从 Handler 注册表查找对应处理器
+3. 按序调用 handler.execute(ctx, config)，收集结果
+4. 异常时记录错误并继续（或中止，取决于配置）
+
+Handler 注册表是 step_key → BaseStepHandler 的映射。
+新增步骤只需：写一个 Handler 类 + 在注册表中加一行。
+"""
+
+from __future__ import annotations
+
+import time
+from pathlib import Path
+from typing import Any, Callable, Dict, List, Optional
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+
+
+class PipelineScheduler:
+    """极简管道调度器。
+
+    用法::
+
+        scheduler = PipelineScheduler(work_dir="./work_dir")
+        scheduler.register_handler(Step1WaterMaskHandler())
+        scheduler.register_handler(Step2GlintDetectionHandler())
+        # ... 注册所有步骤 ...
+
+        scheduler.set_callback(my_callback)  # 可选：GUI 进度回调
+
+        result = scheduler.run_full_pipeline(config)
+        # result['step1'] → {'water_mask_path': ...}
+        # result['step2'] → {'glint_mask_path': ...}
+        # ...
+    """
+
+    def __init__(self, work_dir: str = "./work_dir"):
+        self.ctx = PipelineContext(work_dir)
+        self._handlers: Dict[str, BaseStepHandler] = {}
+
+    # ═══════════════════════════════════════════════════════════
+    # Handler 注册
+    # ═══════════════════════════════════════════════════════════
+
+    def register_handler(self, handler: BaseStepHandler):
+        """注册一个步骤处理器。
+
+        Args:
+            handler: BaseStepHandler 实例（其 step_key 类属性决定 config 中的 key）
+        """
+        if handler.step_key is None:
+            raise ValueError(
+                f"Handler {type(handler).__name__} 未定义 step_key 类属性"
+            )
+        self._handlers[handler.step_key] = handler
+
+    def register_handlers(self, handlers: List[BaseStepHandler]):
+        """批量注册步骤处理器。"""
+        for h in handlers:
+            self.register_handler(h)
+
+    # ═══════════════════════════════════════════════════════════
+    # 回调
+    # ═══════════════════════════════════════════════════════════
+
+    def set_callback(self, callback: Callable):
+        """设置 GUI 进度回调，代理到 PipelineContext。"""
+        self.ctx.set_callback(callback)
+
+    # ═══════════════════════════════════════════════════════════
+    # 单步执行
+    # ═══════════════════════════════════════════════════════════
+
+    def run_step(self, step_key: str, config: dict) -> Dict[str, Any]:
+        """执行单个步骤。
+
+        Args:
+            step_key: 步骤 key（如 'step1', 'step2', ...）
+            config:   该步骤的配置字典
+
+        Returns:
+            步骤执行结果字典
+
+        Raises:
+            KeyError: 如果 step_key 未注册 Handler
+            Exception: 步骤执行中的任何异常
+        """
+        handler = self._handlers.get(step_key)
+        if handler is None:
+            raise KeyError(
+                f"未注册的步骤: {step_key!r}。"
+                f"已注册: {list(self._handlers.keys())}"
+            )
+
+        self.ctx.notify(handler.step_key, 'start')
+        result = handler.execute(self.ctx, config)
+        self.ctx.notify(handler.step_key, 'completed')
+        return result
+
+    # ═══════════════════════════════════════════════════════════
+    # 全流程执行
+    # ═══════════════════════════════════════════════════════════
+
+    def run_full_pipeline(self, config: Dict[str, dict]) -> Dict[str, Any]:
+        """按 config 中的 key 顺序执行全流程。
+
+        遍历 config 的顶层 key，对每个 key：
+        - 如果已注册 Handler → 执行并收集结果
+        - 如果未注册 → 跳过并通知
+        - 如果执行失败 → 记录错误，继续执行后续步骤（不中止）
+
+        Args:
+            config: 全流程配置字典，格式为 {step_key: step_config, ...}
+                    例如: {'step1': {...}, 'step2': {...}, ...}
+
+        Returns:
+            {
+                'step_results': {step_key: result_dict, ...},
+                'step_timings': {...},
+                'total_elapsed': float,
+                'errors': {step_key: error_message, ...},
+            }
+        """
+        self.ctx.pipeline_start_time = time.time()
+
+        step_results: Dict[str, Any] = {}
+        errors: Dict[str, str] = {}
+
+        # 按 config 中的顺序遍历（Python 3.7+ dict 保序）
+        for step_key, step_config in config.items():
+            handler = self._handlers.get(step_key)
+
+            if handler is None:
+                self.ctx.notify(step_key, 'skipped', '未注册 Handler')
+                continue
+
+            try:
+                result = handler.execute(self.ctx, step_config)
+                step_results[step_key] = result
+                self.ctx.notify(step_key, 'completed', str(result))
+            except Exception as e:
+                error_msg = f"{type(e).__name__}: {e}"
+                errors[step_key] = error_msg
+                step_results[step_key] = {'error': error_msg}
+                self.ctx.notify(step_key, 'error', error_msg)
+                # 不中止，继续执行后续步骤
+
+        self.ctx.pipeline_end_time = time.time()
+        total_elapsed = self.ctx.pipeline_end_time - self.ctx.pipeline_start_time
+
+        return {
+            'step_results': step_results,
+            'step_timings': self.ctx.step_timings,
+            'total_elapsed': total_elapsed,
+            'total_elapsed_formatted': self.ctx._format_time(total_elapsed),
+            'errors': errors,
+        }
+
+    # ═══════════════════════════════════════════════════════════
+    # 便捷属性（代理到 PipelineContext）
+    # ═══════════════════════════════════════════════════════════
+
+    @property
+    def work_dir(self) -> Path:
+        return self.ctx.work_dir
+
+    @property
+    def water_mask_path(self) -> Optional[str]:
+        return self.ctx.water_mask_path
+
+    @property
+    def glint_mask_path(self) -> Optional[str]:
+        return self.ctx.glint_mask_path
+
+    @property
+    def deglint_img_path(self) -> Optional[str]:
+        return self.ctx.deglint_img_path
+
+    @property
+    def processed_csv_path(self) -> Optional[str]:
+        return self.ctx.processed_csv_path
+
+    @property
+    def training_csv_path(self) -> Optional[str]:
+        return self.ctx.training_csv_path
+
+    @property
+    def indices_path(self) -> Optional[str]:
+        return self.ctx.indices_path
+
+    def get_step_output_dir(self, step_name: str) -> Path:
+        return self.ctx.get_step_output_dir(step_name)

src/core/handlers/register_handlers.py

@@ -0,0 +1,57 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Handler 注册辅助函数
+
+将所有步骤 Handler 一次性注册到 PipelineScheduler。
+新增步骤只需在此函数中加一行 register_handler() 调用。
+"""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+from src.core.handlers.step1_water_mask import Step1WaterMaskHandler
+from src.core.handlers.step2_glint_detection import Step2GlintDetectionHandler
+from src.core.handlers.step3_glint_removal import Step3GlintRemovalHandler
+from src.core.handlers.step4_sampling import Step4SamplingHandler
+from src.core.handlers.step5_process_csv import Step5ProcessCsvHandler
+from src.core.handlers.step6_extract_spectra import Step6ExtractSpectraHandler
+from src.core.handlers.step7_calc_indices import Step7CalcIndicesHandler
+from src.core.handlers.step8_ml_train import Step8MlTrainHandler
+from src.core.handlers.step9_ml_predict import Step9MlPredictHandler
+from src.core.handlers.step10_qaa_inversion import Step10QaaInversionHandler
+from src.core.handlers.step11_concentration import Step11ConcentrationHandler
+from src.core.handlers.step12_kriging import Step12KrigingHandler
+from src.core.handlers.step13_visualization import Step13VisualizationHandler
+from src.core.handlers.step14_report import Step14ReportHandler
+
+if TYPE_CHECKING:
+    from src.core.handlers.pipeline_scheduler import PipelineScheduler
+
+
+def register_all_handlers(scheduler: PipelineScheduler):
+    """将所有已实现的步骤 Handler 注册到调度器。
+
+    用法::
+
+        scheduler = PipelineScheduler(work_dir="./work_dir")
+        register_all_handlers(scheduler)
+        result = scheduler.run_full_pipeline(config)
+
+    新增步骤时，在此函数中追加一行 register_handler() 即可。
+    """
+    scheduler.register_handler(Step1WaterMaskHandler())
+    scheduler.register_handler(Step2GlintDetectionHandler())
+    scheduler.register_handler(Step3GlintRemovalHandler())
+    scheduler.register_handler(Step4SamplingHandler())
+    scheduler.register_handler(Step5ProcessCsvHandler())
+    scheduler.register_handler(Step6ExtractSpectraHandler())
+    scheduler.register_handler(Step7CalcIndicesHandler())
+    scheduler.register_handler(Step8MlTrainHandler())
+    scheduler.register_handler(Step9MlPredictHandler())
+    scheduler.register_handler(Step10QaaInversionHandler())
+    scheduler.register_handler(Step11ConcentrationHandler())
+    scheduler.register_handler(Step12KrigingHandler())
+    scheduler.register_handler(Step13VisualizationHandler())
+    scheduler.register_handler(Step14ReportHandler())

src/core/handlers/step10_qaa_inversion.py

@@ -0,0 +1,137 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step10 处理器：QAA 准解析算法反演
+
+将原 WaterQualityInversionPipeline.step8_qaa_inversion() 方法
+剥离为独立的 Step10QaaInversionHandler。
+"""
+
+import os
+import time
+from typing import Any, Dict
+
+import numpy as np
+import pandas as pd
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+
+
+class Step10QaaInversionHandler(BaseStepHandler):
+    """步骤10：QAA 准解析算法反演（非经验模型）。
+
+    对应 config key: 'step10_qaa'
+    直接使用 QAABaselineSolver 进行物理推导。
+    """
+
+    step_key = 'step10_qaa'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        from src.core.algorithms.qaa.qaas_baseline import QAABaselineSolver
+        from src.utils.water_owt_config import get_lambda_0
+
+        step_start_time = time.time()
+
+        lake_name = config.get('lake_name', 'Unknown')
+        lambda_0 = config.get('lambda_0', get_lambda_0(lake_name))
+        output_dir = os.path.join(context.work_dir, "10_QAA_Inversion")
+        os.makedirs(output_dir, exist_ok=True)
+        output_path = config.get('output_path') or os.path.join(output_dir, "a_lambda_results.csv")
+
+        spectrum_csv = config.get('spectrum_csv_path')
+        if not spectrum_csv:
+            spectrum_csv = context.training_csv_path
+        if not spectrum_csv or not os.path.exists(spectrum_csv):
+            fallback_candidates = []
+            step6_dir = os.path.join(context.work_dir, "6_Spectral_Feature_Extraction")
+            if os.path.isdir(step6_dir):
+                for f in sorted(os.listdir(step6_dir)):
+                    if f.lower().endswith('.csv'):
+                        fallback_candidates.append(os.path.join(step6_dir, f))
+            if fallback_candidates:
+                spectrum_csv = fallback_candidates[0]
+                context.notify('step10_qaa', 'info',
+                               f'spectrum_csv_path 为空，已自动回退到 step6 产物: {spectrum_csv}')
+            else:
+                msg = f'训练光谱 CSV 不存在或路径为空: {spectrum_csv}'
+                context.notify('step10_qaa', 'error', msg)
+                step_end_time = time.time()
+                context.record_step_time(
+                    "步骤10: QAA 反演", step_start_time, step_end_time,
+                    status="failed", error=msg
+                )
+                return {'error': msg}
+
+        try:
+            df = pd.read_csv(spectrum_csv, encoding="utf-8-sig")
+            col_names = df.columns.tolist()
+
+            wavelength_col_idx = None
+            for i, col in enumerate(col_names):
+                try:
+                    float(col)
+                    wavelength_col_idx = i
+                    break
+                except (ValueError, TypeError):
+                    pass
+
+            if wavelength_col_idx is None:
+                msg = "无法从 CSV 列名中识别波长信息"
+                context.notify('step10_qaa', 'error', msg)
+                step_end_time = time.time()
+                context.record_step_time(
+                    "步骤10: QAA 反演", step_start_time, step_end_time,
+                    status="failed", error=msg
+                )
+                return {'error': msg}
+
+            meta_df = df.iloc[:, :wavelength_col_idx].copy()
+            wavelengths = np.array([float(c) for c in col_names[wavelength_col_idx:]], dtype=np.float64)
+            data_matrix = df.iloc[:, wavelength_col_idx:].values.astype(np.float64)
+            if data_matrix.ndim == 1:
+                data_matrix = data_matrix[np.newaxis, :]
+
+            solver = QAABaselineSolver()
+            raw_result = solver.run_inversion(wavelengths, data_matrix, lambda_0)
+
+            if isinstance(raw_result, list):
+                sample_results = raw_result
+            else:
+                sample_results = [raw_result]
+
+            rows_out = []
+            for i, sample_result in enumerate(sample_results):
+                wl_arr = wavelengths
+                a_arr = sample_result['a_lambda']
+                bb_arr = sample_result['bb_lambda']
+                meta_row = meta_df.iloc[i].to_dict() if i < len(meta_df) else {}
+                for j, wl in enumerate(wl_arr):
+                    rows_out.append({
+                        'sample_id': f"sample_{i}",
+                        'Wavelength': wl,
+                        'a_lambda': a_arr[j],
+                        'bb_lambda': bb_arr[j],
+                        **meta_row,
+                    })
+
+            result_df = pd.DataFrame(rows_out)
+            result_df.to_csv(output_path, index=False, float_format='%.8f')
+
+            context.qaa_output_path = output_path
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤10: QAA 反演", step_start_time, step_end_time
+            )
+            context.notify('step10_qaa', 'completed',
+                           f"QAA 反演完毕，水域={lake_name}，λ₀={lambda_0}nm")
+
+            return {'qaa_output_path': output_path}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤10: QAA 反演", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/handlers/step11_concentration.py

@@ -0,0 +1,71 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step11 处理器：浓度反演
+
+将原 WaterQualityInversionPipeline.step9_concentration_inversion() 方法
+剥离为独立的 Step11ConcentrationHandler。
+"""
+
+import os
+import time
+from typing import Any, Dict
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+
+
+class Step11ConcentrationHandler(BaseStepHandler):
+    """步骤11：浓度反演（基于 QAA Step10 输出的 a_lambda/bb_lambda）。
+
+    对应 config key: 'step11_concentration'
+    直接使用 ConcentrationPipeline 进行浓度反演。
+    """
+
+    step_key = 'step11_concentration'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        from src.core.algorithms.concentration_inversion import ConcentrationPipeline
+
+        step_start_time = time.time()
+
+        input_csv = config.get('input_csv') or context.qaa_output_path
+        output_csv = config.get('output_csv')
+        lake_case = config.get('lake_case', 'medium')
+
+        if not input_csv or not os.path.exists(input_csv):
+            msg = f"QAA 结果文件不存在或路径为空: {input_csv}"
+            context.notify('step11_concentration', 'error', msg)
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤11: 浓度反演", step_start_time, step_end_time,
+                status="failed", error=msg
+            )
+            return {'error': msg}
+
+        if not output_csv:
+            output_dir = os.path.join(context.work_dir, "11_Concentration")
+            os.makedirs(output_dir, exist_ok=True)
+            output_csv = os.path.join(output_dir, "final_concentrations.csv")
+
+        try:
+            pipeline = ConcentrationPipeline(lake_case=lake_case)
+            result_csv = pipeline.run_pipeline(input_csv, output_csv)
+
+            context.concentration_output_path = result_csv
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤11: 浓度反演", step_start_time, step_end_time
+            )
+            context.notify('step11_concentration', 'completed',
+                           f"浓度反演完毕，结果保存于: {result_csv}")
+
+            return {'concentration_output_path': result_csv}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤11: 浓度反演", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/handlers/step12_kriging.py

@@ -0,0 +1,81 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step12 处理器：克里金空间插值与分布图生成
+
+将原 WaterQualityInversionPipeline.step10_map() 方法
+剥离为独立的 Step12KrigingHandler。
+"""
+
+import time
+from pathlib import Path
+from typing import Any, Dict
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+from src.core.steps.mapping_step import MappingStep
+
+
+class Step12KrigingHandler(BaseStepHandler):
+    """步骤12：克里金空间插值与分布图生成。
+
+    对应 config key: 'step12_kriging'
+    委托类: MappingStep.generate_distribution_map()
+    """
+
+    step_key = 'step12_kriging'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        step_start_time = time.time()
+
+        prediction_csv_path = config.get('prediction_csv_path')
+        boundary_shp_path = config.get('boundary_shp_path')
+
+        # 强制输出到 visualization_dir
+        csv_name = Path(prediction_csv_path).stem if prediction_csv_path else "distribution"
+        forced_image_path = str(context.visualization_dir / f"{csv_name}_distribution.png")
+        viz_dir_resolved = str(context.visualization_dir)
+
+        output_image_path = config.get('output_image_path')
+        if output_image_path and output_image_path != forced_image_path:
+            norm_user = output_image_path.replace('\\', '/').rstrip('/')
+            norm_viz = viz_dir_resolved.replace('\\', '/').rstrip('/')
+            if not norm_user.startswith(norm_viz + '/') and norm_user != norm_viz:
+                output_image_path = forced_image_path
+        else:
+            output_image_path = forced_image_path
+
+        try:
+            result = MappingStep.generate_distribution_map(
+                prediction_csv_path=prediction_csv_path,
+                boundary_shp_path=boundary_shp_path,
+                output_image_path=output_image_path,
+                resolution=config.get('resolution', 30),
+                input_crs=config.get('input_crs', 'EPSG:32651'),
+                output_crs=config.get('output_crs', 'EPSG:4326'),
+                show_sample_points=config.get('show_sample_points', False),
+                base_map_tif=config.get('base_map_tif'),
+                use_distance_diffusion=config.get('use_distance_diffusion', True),
+                max_diffusion_distance=config.get('max_diffusion_distance'),
+                diffusion_power=config.get('diffusion_power', 2),
+                diffusion_n_neighbors=config.get('diffusion_n_neighbors', 15),
+                cmap=config.get('cmap'),
+                expand_ratio=config.get('expand_ratio', 0.05),
+                output_dir=str(context.visualization_dir),
+            )
+
+            context.distribution_map_path = result
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤12: 克里金插值与分布图", step_start_time, step_end_time
+            )
+
+            return {'distribution_map_path': result}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤12: 克里金插值与分布图", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/handlers/step13_visualization.py

@@ -0,0 +1,349 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step13 处理器：可视化成图
+
+将原 WaterQualityInversionPipeline 中的可视化方法
+（散点图、箱型图、光谱曲线、统计图表、耀斑预览）
+剥离为独立的 Step13VisualizationHandler。
+"""
+
+import time
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+import seaborn as sns
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+
+
+class Step13VisualizationHandler(BaseStepHandler):
+    """步骤13：可视化成图。
+
+    对应 config key: 'step13_visualization'
+    包含：散点图、箱型图、光谱曲线、统计图表、耀斑预览。
+    """
+
+    step_key = 'step13_visualization'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        step_start_time = time.time()
+        output_files: Dict[str, Any] = {}
+
+        try:
+            # ── 散点图 ──
+            if config.get('generate_scatter', True):
+                if context.training_csv_path and context.models_dir.exists():
+                    try:
+                        scatter_config = config.get('scatter_config', {})
+                        scatter_paths = self._generate_scatter_plots(context, scatter_config)
+                        output_files['scatter_plots'] = scatter_paths
+                    except Exception as e:
+                        context.notify('step13_visualization', 'warning',
+                                       f"生成散点图时出错: {e}")
+
+            # ── 箱型图 ──
+            if config.get('generate_boxplots', True):
+                if context.processed_csv_path:
+                    try:
+                        boxplot_config = config.get('boxplot_config', {})
+                        boxplot_paths = self._generate_boxplots(context, boxplot_config)
+                        output_files['boxplots'] = boxplot_paths
+                    except Exception as e:
+                        context.notify('step13_visualization', 'warning',
+                                       f"生成箱型图时出错: {e}")
+
+            # ── 光谱曲线 ──
+            if config.get('generate_spectrum', True):
+                if context.training_csv_path:
+                    try:
+                        spectrum_paths = self._generate_spectrum_plots(context, config)
+                        output_files['spectrum_plots'] = spectrum_paths
+                    except Exception as e:
+                        context.notify('step13_visualization', 'warning',
+                                       f"生成光谱曲线图时出错: {e}")
+
+            # ── 统计图表 ──
+            if config.get('generate_statistics', True):
+                if context.processed_csv_path:
+                    try:
+                        stat_charts = self._generate_statistics(context)
+                        output_files['statistical_charts'] = stat_charts
+                    except Exception as e:
+                        context.notify('step13_visualization', 'warning',
+                                       f"生成统计图表时出错: {e}")
+
+            # ── 耀斑预览 ──
+            if config.get('generate_glint_previews', True):
+                try:
+                    glint_config = config.get('glint_preview_config', {})
+                    preview_paths = context.visualizer.generate_glint_deglint_previews(
+                        work_dir=glint_config.get('work_dir') or str(context.work_dir),
+                        output_subdir=glint_config.get('output_subdir', 'glint_deglint_previews'),
+                        generate_glint=glint_config.get('generate_glint', True),
+                        generate_deglint=glint_config.get('generate_deglint', True),
+                    )
+                    output_files['glint_deglint_previews'] = preview_paths
+                except Exception as e:
+                    context.notify('step13_visualization', 'warning',
+                                   f"生成耀斑预览图时出错: {e}")
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤13: 可视化成图", step_start_time, step_end_time
+            )
+
+            return {'visualization_outputs': output_files}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤13: 可视化成图", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise
+
+    # ── 散点图 ──
+
+    def _generate_scatter_plots(self, context: PipelineContext,
+                                 scatter_config: dict) -> Dict[str, str]:
+        training_csv_path = context.training_csv_path
+        models_dir = str(context.models_dir)
+        metric = scatter_config.get('metric', 'test_r2')
+        use_enhanced = scatter_config.get('use_enhanced', True)
+        feature_start_column = scatter_config.get('feature_start_column', 13)
+        test_size = scatter_config.get('test_size', 0.2)
+        random_state = scatter_config.get('random_state', 42)
+
+        scatter_paths = {}
+
+        if use_enhanced:
+            try:
+                results = context.scatter_batch.batch_plot_scatter(
+                    models_root_dir=models_dir,
+                    csv_path=training_csv_path,
+                    output_dir=str(context.visualization_dir / "scatter_plots"),
+                    metric=metric,
+                    target_column=None,
+                    feature_start_column=feature_start_column,
+                    test_size=test_size,
+                    random_state=random_state,
+                )
+                for target_name, result in results.items():
+                    if result.get('status') == 'success':
+                        scatter_paths[target_name] = result.get('save_path', '')
+            except Exception:
+                use_enhanced = False
+
+        if not use_enhanced or not scatter_paths:
+            from src.core.prediction.inference_batch import WaterQualityInference
+            models_path = Path(models_dir)
+            for target_folder in models_path.iterdir():
+                if not target_folder.is_dir():
+                    continue
+                target_name = target_folder.name
+                try:
+                    inferencer = WaterQualityInference(str(target_folder))
+                    eval_result = inferencer.evaluate_with_split(
+                        data_csv_path=training_csv_path,
+                        split_method="spxy",
+                        test_size=test_size,
+                        random_state=random_state,
+                        metric=metric,
+                    )
+                    predictions = eval_result.get('predictions', {})
+                    if predictions:
+                        y_train_true = predictions.get('y_train_true')
+                        y_train_pred = predictions.get('y_train_pred')
+                        y_test_true = predictions.get('y_test_true')
+                        y_test_pred = predictions.get('y_test_pred')
+                        metrics = eval_result.get('test_metrics', {})
+                        if y_train_true is not None and y_test_true is not None:
+                            y_all_true = np.concatenate([y_train_true, y_test_true])
+                            y_all_pred = np.concatenate([y_train_pred, y_test_pred])
+                            train_indices = np.arange(len(y_train_true))
+                            test_indices = np.arange(len(y_train_true), len(y_all_true))
+                            scatter_path = context.visualizer.plot_scatter_true_vs_pred(
+                                y_true=y_all_true,
+                                y_pred=y_all_pred,
+                                target_name=target_name,
+                                train_indices=train_indices,
+                                test_indices=test_indices,
+                                metrics={
+                                    'train_r2': eval_result.get('train_metrics', {}).get('r2', 0),
+                                    'test_r2': metrics.get('r2', 0),
+                                    'train_rmse': eval_result.get('train_metrics', {}).get('rmse', 0),
+                                    'test_rmse': metrics.get('rmse', 0),
+                                }
+                            )
+                            scatter_paths[target_name] = scatter_path
+                except Exception:
+                    continue
+
+        return scatter_paths
+
+    # ── 箱型图 ──
+
+    def _generate_boxplots(self, context: PipelineContext,
+                            boxplot_config: dict) -> Dict[str, str]:
+        csv_path = context.processed_csv_path
+        parameter_columns = boxplot_config.get('parameter_columns')
+        data_start_column = boxplot_config.get('data_start_column', 4)
+        save_individual = boxplot_config.get('save_individual', True)
+        use_seaborn = boxplot_config.get('use_seaborn', True)
+
+        df = pd.read_csv(csv_path)
+
+        if parameter_columns is None:
+            data_columns = df.iloc[:, data_start_column:]
+            parameter_columns = list(data_columns.columns)
+        else:
+            parameter_columns = [col for col in parameter_columns if col in df.columns]
+
+        if not parameter_columns:
+            return {}
+
+        boxplot_dir = context.visualization_dir / "boxplots"
+        boxplot_dir.mkdir(parents=True, exist_ok=True)
+        boxplot_paths = {}
+
+        if save_individual:
+            for column in parameter_columns:
+                if column not in df.columns:
+                    continue
+                clean_data = df[column].dropna()
+                if len(clean_data) == 0:
+                    continue
+                try:
+                    plt.figure(figsize=(8, 6))
+                    if use_seaborn:
+                        plot_data = pd.DataFrame({'参数': [column] * len(clean_data), '数值': clean_data})
+                        sns.boxplot(data=plot_data, x='参数', y='数值', palette='Set2')
+                        sns.stripplot(data=plot_data, x='参数', y='数值',
+                                     color='red', alpha=0.6, size=5, jitter=True)
+                    else:
+                        box_plot = plt.boxplot([clean_data], labels=[column],
+                                              patch_artist=True, showfliers=False)
+                        box_plot['boxes'][0].set_facecolor('lightblue')
+                        box_plot['boxes'][0].set_alpha(0.7)
+                        x_pos = np.random.normal(1, 0.04, size=len(clean_data))
+                        plt.scatter(x_pos, clean_data, alpha=0.6, s=30, color='red',
+                                   edgecolors='black', linewidth=0.5, zorder=3)
+                    plt.title(f'{column} - 箱型图', fontsize=14, fontweight='bold')
+                    plt.xlabel('参数', fontsize=12)
+                    plt.ylabel('数值', fontsize=12)
+                    stats_text = (f'数据点数: {len(clean_data)}\n'
+                                f'均值: {clean_data.mean():.2f}\n'
+                                f'中位数: {clean_data.median():.2f}\n'
+                                f'标准差: {clean_data.std():.2f}')
+                    plt.text(0.02, 0.98, stats_text, transform=plt.gca().transAxes,
+                            verticalalignment='top',
+                            bbox=dict(boxstyle='round',
+                                    facecolor='wheat' if not use_seaborn else 'lightgreen',
+                                    alpha=0.8))
+                    plt.grid(True, alpha=0.3, linestyle='--')
+                    plt.tight_layout()
+                    safe_name = column.replace('/', '_').replace('\\', '_').replace(':', '_')
+                    save_path = boxplot_dir / f'{safe_name}_boxplot.png'
+                    plt.savefig(save_path, dpi=300, bbox_inches='tight')
+                    plt.close()
+                    boxplot_paths[column] = str(save_path)
+                except Exception:
+                    continue
+
+        # 综合箱型图
+        try:
+            plt.figure(figsize=(max(12, len(parameter_columns) * 0.8), 8))
+            box_data = []
+            labels = []
+            for column in parameter_columns:
+                if column in df.columns:
+                    clean_data = df[column].dropna()
+                    if len(clean_data) > 0:
+                        box_data.append(clean_data)
+                        labels.append(column)
+            if box_data:
+                if use_seaborn:
+                    melted_data = pd.melt(df[labels], var_name='参数', value_name='数值')
+                    melted_data = melted_data.dropna()
+                    sns.boxplot(data=melted_data, x='参数', y='数值', palette='Set3')
+                    sns.stripplot(data=melted_data, x='参数', y='数值',
+                                 color='red', alpha=0.6, size=4, jitter=True)
+                else:
+                    box_plot = plt.boxplot(box_data, labels=labels, patch_artist=True, showfliers=False)
+                    colors = plt.cm.Set3(np.linspace(0, 1, len(box_data)))
+                    for patch, color in zip(box_plot['boxes'], colors):
+                        patch.set_facecolor(color)
+                        patch.set_alpha(0.7)
+                    for i, data in enumerate(box_data):
+                        x_pos = np.random.normal(i + 1, 0.04, size=len(data))
+                        plt.scatter(x_pos, data, alpha=0.6, s=20, color='red',
+                                   edgecolors='black', linewidth=0.5, zorder=3)
+                plt.title('水质参数箱型图（综合）', fontsize=16, fontweight='bold')
+                plt.xlabel('参数', fontsize=12)
+                plt.ylabel('数值', fontsize=12)
+                plt.xticks(rotation=45, ha='right')
+                plt.grid(True, alpha=0.3, linestyle='--')
+                plt.tight_layout()
+                combined_path = boxplot_dir / 'all_parameters_boxplot.png'
+                plt.savefig(combined_path, dpi=300, bbox_inches='tight')
+                plt.close()
+                boxplot_paths['all_parameters'] = str(combined_path)
+        except Exception:
+            pass
+
+        return boxplot_paths
+
+    # ── 光谱曲线 ──
+
+    def _generate_spectrum_plots(self, context: PipelineContext,
+                                  config: dict) -> Dict[str, str]:
+        csv_path = context.training_csv_path
+        wavelength_start_column = config.get('feature_start_column', 'UTM_Y')
+
+        df = pd.read_csv(csv_path)
+        if isinstance(wavelength_start_column, str):
+            try:
+                wavelength_start_idx = df.columns.get_loc(wavelength_start_column)
+            except KeyError:
+                wavelength_start_idx = 13
+        else:
+            wavelength_start_idx = wavelength_start_column
+
+        parameter_columns = list(df.columns[:wavelength_start_idx])
+        if len(parameter_columns) > 2:
+            parameter_columns = parameter_columns[2:]
+
+        spectrum_paths = {}
+        for param_col in parameter_columns:
+            if param_col not in df.columns:
+                continue
+            try:
+                spectrum_path = context.visualizer.plot_spectrum_by_parameter(
+                    csv_path=csv_path,
+                    parameter_column=param_col,
+                    wavelength_start_column=wavelength_start_column,
+                    n_groups=5,
+                )
+                spectrum_paths[param_col] = spectrum_path
+            except Exception:
+                continue
+
+        return spectrum_paths
+
+    # ── 统计图表 ──
+
+    def _generate_statistics(self, context: PipelineContext) -> Dict[str, str]:
+        csv_path = context.processed_csv_path
+        df = pd.read_csv(csv_path)
+        parameter_columns = list(df.columns[2:])
+        parameter_columns = [col for col in parameter_columns
+                           if df[col].dtype in [np.float64, np.int64]]
+
+        return context.visualizer.plot_statistical_charts(
+            csv_path=csv_path,
+            parameter_columns=parameter_columns,
+        )

src/core/handlers/step14_report.py

@@ -0,0 +1,142 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step14 处理器：报告生成
+
+将原 WaterQualityInversionPipeline.generate_pipeline_report() 方法
+剥离为独立的 Step14ReportHandler。
+"""
+
+import time
+from datetime import datetime
+from pathlib import Path
+from typing import Any, Dict
+
+import numpy as np
+import pandas as pd
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+
+
+class Step14ReportHandler(BaseStepHandler):
+    """步骤14：流程执行报告生成。
+
+    对应 config key: 'step14_report'
+    生成 CSV 和 TXT 格式的流程执行报告。
+    """
+
+    step_key = 'step14_report'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        step_start_time = time.time()
+
+        try:
+            output_path = config.get('output_path')
+            if output_path is None:
+                timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
+                output_path = str(context.reports_dir / f"pipeline_report_{timestamp}.csv")
+
+            report_data = []
+            total_time = 0.0
+
+            step_order = [
+                "步骤1: 水域掩膜生成",
+                "步骤2: 耀斑区域检测",
+                "步骤3: 耀斑去除",
+                "步骤4: 数据预处理",
+                "步骤5: 光谱提取",
+                "步骤6: 水质光谱指数计算",
+                "步骤7: 机器学习建模与训练",
+                "步骤8: 非经验模型训练",
+                "步骤9: 自定义回归",
+                "步骤10: 采样点生成",
+                "步骤11: 参数预测",
+                "步骤12: 分布图生成",
+            ]
+
+            for step_name in step_order:
+                if step_name in context.step_timings:
+                    timing_info = context.step_timings[step_name]
+                    report_data.append({
+                        '步骤': step_name,
+                        '开始时间': timing_info['start_time'],
+                        '结束时间': timing_info['end_time'],
+                        '耗时(秒)': f"{timing_info['elapsed_seconds']:.2f}",
+                        '耗时(格式化)': timing_info['elapsed_formatted'],
+                        '状态': timing_info['status'],
+                        '错误信息': timing_info.get('error', '')
+                    })
+                    if timing_info['status'] == 'completed':
+                        total_time += timing_info['elapsed_seconds']
+
+            if context.pipeline_start_time and context.pipeline_end_time:
+                pipeline_total = context.pipeline_end_time - context.pipeline_start_time
+                report_data.append({
+                    '步骤': '总计',
+                    '开始时间': datetime.fromtimestamp(context.pipeline_start_time).strftime('%Y-%m-%d %H:%M:%S'),
+                    '结束时间': datetime.fromtimestamp(context.pipeline_end_time).strftime('%Y-%m-%d %H:%M:%S'),
+                    '耗时(秒)': f"{pipeline_total:.2f}",
+                    '耗时(格式化)': context._format_time(pipeline_total),
+                    '状态': 'completed',
+                    '错误信息': ''
+                })
+
+            df_report = pd.DataFrame(report_data)
+            df_report.to_csv(output_path, index=False, encoding='utf-8-sig')
+
+            txt_output_path = str(Path(output_path).with_suffix('.txt'))
+            with open(txt_output_path, 'w', encoding='utf-8') as f:
+                f.write("=" * 80 + "\n")
+                f.write("水质参数反演流程执行报告\n")
+                f.write("=" * 80 + "\n\n")
+
+                if context.pipeline_start_time and context.pipeline_end_time:
+                    f.write(f"流程开始时间: {datetime.fromtimestamp(context.pipeline_start_time).strftime('%Y-%m-%d %H:%M:%S')}\n")
+                    f.write(f"流程结束时间: {datetime.fromtimestamp(context.pipeline_end_time).strftime('%Y-%m-%d %H:%M:%S')}\n")
+                    f.write(f"总耗时: {context._format_time(context.pipeline_end_time - context.pipeline_start_time)}\n\n")
+
+                f.write("-" * 80 + "\n")
+                f.write("各步骤执行详情:\n")
+                f.write("-" * 80 + "\n\n")
+
+                for step_name in step_order:
+                    if step_name in context.step_timings:
+                        timing_info = context.step_timings[step_name]
+                        f.write(f"{step_name}\n")
+                        f.write(f"  开始时间: {timing_info['start_time']}\n")
+                        f.write(f"  结束时间: {timing_info['end_time']}\n")
+                        f.write(f"  耗时: {timing_info['elapsed_formatted']} ({timing_info['elapsed_seconds']:.2f}秒)\n")
+                        f.write(f"  状态: {timing_info['status']}\n")
+                        if timing_info.get('error'):
+                            f.write(f"  错误: {timing_info['error']}\n")
+                        f.write("\n")
+
+                f.write("-" * 80 + "\n")
+                f.write("统计摘要:\n")
+                f.write("-" * 80 + "\n")
+                completed_steps = [s for s in context.step_timings.values() if s['status'] == 'completed']
+                failed_steps = [s for s in context.step_timings.values() if s['status'] == 'failed']
+                skipped_steps = [s for s in context.step_timings.values() if s['status'] == 'skipped']
+                f.write(f"成功完成的步骤: {len(completed_steps)}\n")
+                f.write(f"失败的步骤: {len(failed_steps)}\n")
+                f.write(f"跳过的步骤: {len(skipped_steps)}\n")
+                if completed_steps:
+                    completed_times = [s['elapsed_seconds'] for s in completed_steps]
+                    f.write(f"平均耗时: {context._format_time(np.mean(completed_times))}\n")
+                    f.write(f"最长耗时: {context._format_time(np.max(completed_times))}\n")
+                    f.write(f"最短耗时: {context._format_time(np.min(completed_times))}\n")
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤14: 报告生成", step_start_time, step_end_time
+            )
+
+            return {'report_csv': output_path, 'report_txt': txt_output_path}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤14: 报告生成", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/handlers/step1_water_mask.py

@@ -0,0 +1,83 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step1 处理器：水域掩膜生成
+
+将原 WaterQualityInversionPipeline.step1_generate_water_mask() 方法
+剥离为独立的 Step1WaterMaskHandler。
+
+这是 14 个步骤 Handler 的**打样模板**，其余步骤照此模式拆分：
+1. 继承 BaseStepHandler，设置 step_key 类属性
+2. 实现 execute(ctx, config) → 调用对应 Step 类的静态方法
+3. 将输出路径写入 ctx（上下文共享）
+4. 记录步骤耗时
+5. 返回结果字典
+"""
+
+import time
+from typing import Any, Dict
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+from src.core.steps.water_mask_step import WaterMaskStep
+
+
+class Step1WaterMaskHandler(BaseStepHandler):
+    """步骤1：水域掩膜生成。
+
+    对应 config key: 'step1'
+    委托类: WaterMaskStep.run()
+
+    用法::
+
+        handler = Step1WaterMaskHandler()
+        result = handler.execute(ctx, config['step1'])
+        # ctx.water_mask_path 已被更新
+    """
+
+    step_key = 'step1'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        """执行水域掩膜生成。
+
+        config 可包含的键（全部透传给 WaterMaskStep.run()）：
+        - mask_path:     水体掩膜文件路径（.shp / .dat / .tif）
+        - img_path:      输入影像路径（shp 栅格化或 NDWI 时需要）
+        - ndwi_threshold: NDWI 阈值（默认 0.4）
+        - use_ndwi:      是否使用 NDWI 方法（默认 False）
+        - generate_png:  是否生成 PNG 预览（默认 True）
+        - output_path:   指定输出路径（可选）
+
+        Returns:
+            {'water_mask_path': str}
+        """
+        step_start_time = time.time()
+
+        try:
+            result = WaterMaskStep.run(
+                mask_path=config.get('mask_path'),
+                img_path=config.get('img_path'),
+                ndwi_threshold=config.get('ndwi_threshold', 0.4),
+                use_ndwi=config.get('use_ndwi', False),
+                generate_png=config.get('generate_png', True),
+                output_path=config.get('output_path'),
+                water_mask_dir=str(context.water_mask_dir),
+                callback=context.notify,
+            )
+
+            # 将输出路径写入上下文，供后续步骤使用
+            context.water_mask_path = result
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤1: 水域掩膜生成", step_start_time, step_end_time
+            )
+
+            return {'water_mask_path': result}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤1: 水域掩膜生成", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/handlers/step2_glint_detection.py

@@ -0,0 +1,67 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step2 处理器：耀斑区域检测
+
+将原 WaterQualityInversionPipeline.step2_find_glint_area() 方法
+剥离为独立的 Step2GlintDetectionHandler。
+"""
+
+import time
+from typing import Any, Dict
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+from src.core.steps.glint_detection_step import GlintDetectionStep
+
+
+class Step2GlintDetectionHandler(BaseStepHandler):
+    """步骤2：耀斑区域检测。
+
+    对应 config key: 'step2'
+    委托类: GlintDetectionStep.run()
+    """
+
+    step_key = 'step2'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        step_start_time = time.time()
+
+        water_mask_path = self._resolve_path(
+            config.get('water_mask_path'), context.water_mask_path, 'water_mask'
+        )
+
+        try:
+            result = GlintDetectionStep.run(
+                img_path=config.get('img_path'),
+                glint_wave=config.get('glint_wave', 750.0),
+                method=config.get('method', 'otsu'),
+                z_threshold=config.get('z_threshold', 2.5),
+                percentile=config.get('percentile', 95.0),
+                iqr_multiplier=config.get('iqr_multiplier', 1.5),
+                window_size=config.get('window_size', 15),
+                multi_band_waves=config.get('multi_band_waves'),
+                sub_method=config.get('sub_method', 'zscore'),
+                weights=config.get('weights'),
+                max_area=config.get('max_area'),
+                buffer_size=config.get('buffer_size'),
+                water_mask_path=water_mask_path,
+                glint_dir=str(context.glint_dir),
+                callback=context.notify,
+            )
+
+            context.glint_mask_path = result
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤2: 耀斑区域检测", step_start_time, step_end_time
+            )
+
+            return {'glint_mask_path': result}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤2: 耀斑区域检测", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/handlers/step3_glint_removal.py

@@ -0,0 +1,85 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step3 处理器：耀斑去除
+
+将原 WaterQualityInversionPipeline.step3_remove_glint() 方法
+剥离为独立的 Step3GlintRemovalHandler。
+"""
+
+import time
+from typing import Any, Dict
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+from src.core.steps.glint_removal_step import GlintRemovalStep
+
+
+class Step3GlintRemovalHandler(BaseStepHandler):
+    """步骤3：耀斑去除。
+
+    对应 config key: 'step3'
+    委托类: GlintRemovalStep.run()
+    """
+
+    step_key = 'step3'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        step_start_time = time.time()
+
+        water_mask_path = self._resolve_path(
+            config.get('water_mask_path'), context.water_mask_path, 'water_mask'
+        )
+
+        try:
+            result = GlintRemovalStep.run(
+                img_path=config.get('img_path'),
+                method=config.get('method', 'subtract_nir'),
+                start_wave=config.get('start_wave'),
+                end_wave=config.get('end_wave'),
+                json_path=config.get('json_path'),
+                left_shoulder_wave=config.get('left_shoulder_wave'),
+                valley_wave=config.get('valley_wave'),
+                right_shoulder_wave=config.get('right_shoulder_wave'),
+                water_mask=water_mask_path,
+                interpolate_zeros=config.get('interpolate_zeros', False),
+                interpolation_method=config.get('interpolation_method', 'nearest'),
+                enabled=config.get('enabled', True),
+                kutser_shp_path=config.get('kutser_shp_path'),
+                oxy_band=config.get('oxy_band', 38),
+                lower_oxy=config.get('lower_oxy', 36),
+                upper_oxy=config.get('upper_oxy', 49),
+                nir_band=config.get('nir_band', 47),
+                nir_lower=config.get('nir_lower', 25),
+                nir_upper=config.get('nir_upper', 37),
+                goodman_A=config.get('goodman_A', 0.000019),
+                goodman_B=config.get('goodman_B', 0.1),
+                hedley_shp_path=config.get('hedley_shp_path'),
+                hedley_nir_band=config.get('hedley_nir_band', 47),
+                sugar_bounds=config.get('sugar_bounds'),
+                sugar_sigma=config.get('sugar_sigma', 1.0),
+                sugar_estimate_background=config.get('sugar_estimate_background', True),
+                sugar_glint_mask_method=config.get('sugar_glint_mask_method', 'cdf'),
+                sugar_iter=config.get('sugar_iter', 3),
+                sugar_termination_thresh=config.get('sugar_termination_thresh', 20.0),
+                deglint_dir=str(context.deglint_dir),
+                water_mask_dir=str(context.water_mask_dir),
+                callback=context.notify,
+                output_path=config.get('output_path'),
+            )
+
+            context.deglint_img_path = result
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤3: 耀斑去除", step_start_time, step_end_time
+            )
+
+            return {'deglint_img_path': result}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤3: 耀斑去除", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/handlers/step4_sampling.py

@@ -0,0 +1,64 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step4 处理器：预测采样点生成
+
+将原 WaterQualityInversionPipeline.step4_sampling() 方法
+剥离为独立的 Step4SamplingHandler。
+"""
+
+import time
+from typing import Any, Dict
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+from src.core.steps.prediction_step import PredictionStep
+
+
+class Step4SamplingHandler(BaseStepHandler):
+    """步骤4：生成预测采样点并提取光谱。
+
+    对应 config key: 'step4_sampling'
+    委托类: PredictionStep.generate_sampling_points()
+    """
+
+    step_key = 'step4_sampling'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        step_start_time = time.time()
+
+        deglint_img_path = self._resolve_path(
+            config.get('deglint_img_path'), context.deglint_img_path, 'deglint_img'
+        )
+        water_mask_path = self._resolve_path(
+            config.get('water_mask_path'), context.water_mask_path, 'water_mask'
+        )
+        glint_mask_path = self._resolve_path(
+            config.get('glint_mask_path'), context.glint_mask_path, 'glint_mask'
+        )
+
+        try:
+            result = PredictionStep.generate_sampling_points(
+                deglint_img_path=deglint_img_path,
+                interval=config.get('interval', 50),
+                sample_radius=config.get('sample_radius', 5),
+                chunk_size=config.get('chunk_size', 1000),
+                water_mask_path=water_mask_path,
+                glint_mask_path=glint_mask_path,
+                output_dir=str(context.sampling_dir),
+                use_adaptive_sampling=config.get('use_adaptive_sampling', True),
+            )
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤4: 生成预测采样点", step_start_time, step_end_time
+            )
+
+            return {'sampling_csv_path': result}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤4: 生成预测采样点", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/handlers/step5_process_csv.py

@@ -0,0 +1,50 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step5 处理器：CSV 数据处理
+
+将原 WaterQualityInversionPipeline.step5_process_csv() 方法
+剥离为独立的 Step5ProcessCsvHandler。
+"""
+
+import time
+from typing import Any, Dict
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+from src.core.steps.data_preparation_step import DataPreparationStep
+
+
+class Step5ProcessCsvHandler(BaseStepHandler):
+    """步骤5：处理 CSV 文件，筛选剔除异常值。
+
+    对应 config key: 'step5_clean'
+    委托类: DataPreparationStep.process_csv()
+    """
+
+    step_key = 'step5_clean'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        step_start_time = time.time()
+
+        try:
+            result = DataPreparationStep.process_csv(
+                csv_path=config.get('csv_path'),
+                output_dir=str(context.processed_data_dir),
+            )
+
+            context.processed_csv_path = result
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤5: 处理CSV文件", step_start_time, step_end_time
+            )
+
+            return {'processed_csv_path': result}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤5: 处理CSV文件", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/handlers/step6_extract_spectra.py

@@ -0,0 +1,66 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step6 处理器：训练样本点光谱提取
+
+将原 WaterQualityInversionPipeline.step6_extract_spectra() 方法
+剥离为独立的 Step6ExtractSpectraHandler。
+"""
+
+import time
+from typing import Any, Dict
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+from src.core.steps.data_preparation_step import DataPreparationStep
+
+
+class Step6ExtractSpectraHandler(BaseStepHandler):
+    """步骤6：根据采样点坐标在去耀斑影像中提取平均光谱。
+
+    对应 config key: 'step6_feature'
+    委托类: DataPreparationStep.extract_training_spectra()
+    """
+
+    step_key = 'step6_feature'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        step_start_time = time.time()
+
+        deglint_img_path = self._resolve_path(
+            config.get('deglint_img_path'), context.deglint_img_path, 'deglint_img'
+        )
+        csv_path = self._resolve_path(
+            config.get('csv_path'), context.processed_csv_path, 'csv'
+        )
+        glint_mask_path = self._resolve_path(
+            config.get('glint_mask_path'), context.glint_mask_path, 'glint_mask'
+        )
+
+        try:
+            result = DataPreparationStep.extract_training_spectra(
+                deglint_img_path=deglint_img_path,
+                radius=config.get('radius', 5),
+                source_epsg=config.get('source_epsg', 4326),
+                csv_path=csv_path,
+                boundary_path=config.get('boundary_path'),
+                glint_mask_path=glint_mask_path,
+                water_mask_path=context.water_mask_path,
+                output_dir=str(context.training_spectra_dir),
+            )
+
+            context.training_csv_path = result
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤6: 提取训练样本点光谱", step_start_time, step_end_time
+            )
+
+            return {'training_csv_path': result}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤6: 提取训练样本点光谱", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/handlers/step7_calc_indices.py

@@ -0,0 +1,58 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step7 处理器：水质光谱指数计算
+
+将原 WaterQualityInversionPipeline.step7_calc_indices() 方法
+剥离为独立的 Step7CalcIndicesHandler。
+"""
+
+import time
+from typing import Any, Dict
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+from src.core.steps.data_preparation_step import DataPreparationStep
+
+
+class Step7CalcIndicesHandler(BaseStepHandler):
+    """步骤7：根据训练光谱计算水质光谱指数。
+
+    对应 config key: 'step7_index'
+    委托类: DataPreparationStep.calculate_water_quality_indices()
+    """
+
+    step_key = 'step7_index'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        step_start_time = time.time()
+
+        training_csv_path = self._resolve_path(
+            config.get('training_csv_path'), context.training_csv_path, 'training_csv'
+        )
+
+        try:
+            result = DataPreparationStep.calculate_water_quality_indices(
+                training_csv_path=training_csv_path,
+                formula_csv_file=config.get('formula_csv_file'),
+                formula_names=config.get('formula_names'),
+                output_file=config.get('output_file'),
+                enabled=config.get('enabled', True),
+                output_dir=str(context.indices_dir),
+            )
+
+            context.indices_path = result
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤7: 计算水质光谱指数", step_start_time, step_end_time
+            )
+
+            return {'indices_path': result}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤7: 计算水质光谱指数", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/handlers/step8_ml_train.py

@@ -0,0 +1,58 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step8 处理器：机器学习建模与训练
+
+将原 WaterQualityInversionPipeline.step8_train_ml() 方法
+剥离为独立的 Step8MlTrainHandler。
+"""
+
+import time
+from typing import Any, Dict
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+from src.core.steps.modeling_step import ModelingStep
+
+
+class Step8MlTrainHandler(BaseStepHandler):
+    """步骤8：机器学习建模与训练。
+
+    对应 config key: 'step8_ml_train'
+    委托类: ModelingStep.train_models()
+    """
+
+    step_key = 'step8_ml_train'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        step_start_time = time.time()
+
+        training_csv_path = self._resolve_path(
+            config.get('training_csv_path'), context.training_csv_path, 'training_csv'
+        )
+
+        try:
+            result = ModelingStep.train_models(
+                feature_start_column=config.get('feature_start_column', '374.285004'),
+                preprocessing_methods=config.get('preprocessing_methods'),
+                model_names=config.get('model_names'),
+                split_methods=config.get('split_methods'),
+                cv_folds=config.get('cv_folds', 5),
+                training_csv_path=training_csv_path,
+                output_dir=str(context.models_dir),
+                _report_generator=context.report_generator,
+            )
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤8: 机器学习建模与训练", step_start_time, step_end_time
+            )
+
+            return {'models_dir': result}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤8: 机器学习建模与训练", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/handlers/step9_ml_predict.py

@@ -0,0 +1,64 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Step9 处理器：机器学习推理预测
+
+将原 WaterQualityInversionPipeline.step9_predict_ml() 方法
+剥离为独立的 Step9MlPredictHandler。
+"""
+
+import time
+from typing import Any, Dict
+
+from src.core.handlers.base import BaseStepHandler, PipelineContext
+from src.core.steps.prediction_step import PredictionStep
+
+
+class Step9MlPredictHandler(BaseStepHandler):
+    """步骤9：机器学习推理预测。
+
+    对应 config key: 'step9_ml_predict'
+    委托类: PredictionStep.predict_water_quality()
+    """
+
+    step_key = 'step9_ml_predict'
+
+    def execute(self, context: PipelineContext, config: dict) -> Dict[str, Any]:
+        step_start_time = time.time()
+
+        sampling_csv_path = self._resolve_path(
+            config.get('sampling_csv_path'), context.sampling_csv_path, 'sampling_csv'
+        )
+
+        models_dir = config.get('models_dir') or str(context.models_dir)
+
+        try:
+            result = PredictionStep.predict_water_quality(
+                sampling_csv_path=sampling_csv_path,
+                models_dir=models_dir,
+                metric=config.get('metric', 'test_r2'),
+                prediction_column=config.get('prediction_column', 'prediction'),
+                output_dir=str(context.prediction_dir / "9_ML_Prediction"),
+                _report_generator=context.report_generator,
+                _external_model=config.get('_external_model'),
+                _external_model_path=config.get('_external_model_path'),
+                _external_models_dict=config.get('_external_models_dict'),
+                _external_model_dir=config.get('_external_model_dir'),
+            )
+
+            context.prediction_files.update(result)
+
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤9: 机器学习推理预测", step_start_time, step_end_time
+            )
+
+            return {'prediction_files': result}
+
+        except Exception as e:
+            step_end_time = time.time()
+            context.record_step_time(
+                "步骤9: 机器学习推理预测", step_start_time, step_end_time,
+                status="failed", error=str(e)
+            )
+            raise

src/core/modeling/modeling_batch.py

@@ -13,7 +13,7 @@ from sklearn.svm import SVR
 from sklearn.ensemble import RandomForestRegressor
 from sklearn.neighbors import KNeighborsRegressor
 from sklearn.linear_model import LinearRegression, Ridge, Lasso, ElasticNet
-from sklearn.model_selection import GridSearchCV, cross_val_score, KFold, train_test_split
+from sklearn.model_selection import RandomizedSearchCV, cross_val_score, KFold, train_test_split
 from sklearn.metrics import mean_squared_error, mean_absolute_error, r2_score
 from sklearn.cross_decomposition import PLSRegression
 from sklearn.ensemble import GradientBoostingRegressor, AdaBoostRegressor, ExtraTreesRegressor
@@ -45,6 +45,7 @@ is_frozen_env = getattr(sys, 'frozen', False)
 safe_n_jobs = 1 if is_frozen_env else -1
 
 from src.preprocessing.spectral_Preprocessing import Preprocessing
+from src.core.utils.split_methods import spxy, ks
 
 
 class WaterQualityModelingBatch:
@@ -288,11 +289,24 @@ class WaterQualityModelingBatch:
         # 提取所有目标列（从0列到feature_start_index-1列）
         y_dict = {}
         target_columns = data.columns[:feature_start_index]
-        
-        print(f"检测到的目标列: {list(target_columns)}")
-        
+        print(f"检测到的潜在目标列: {list(target_columns)}")
+
+        # 新增：跳过非预测目标的系统保留列
+        ignore_cols = {'ID', 'id', 'Id', 'Longitude', 'Latitude', 'Lon', 'Lat', 'longitude', 'latitude', 'lon', 'lat', 'Station', 'station'}
+
         for col_name in target_columns:
+            # 过滤黑名单列
+            if col_name in ignore_cols:
+                print(f"  跳过目标列 '{col_name}': 属于系统保留列或空间坐标")
+                continue
+
             y_series = data[col_name]
+
+            # 过滤非数值类型列 (避免将纯文本备注等拿去回归)
+            if not pd.api.types.is_numeric_dtype(y_series):
+                print(f"  跳过目标列 '{col_name}': 非数值类型")
+                continue
+
             # 检查是否有非空值
             if not y_series.isna().all():
                 y_dict[col_name] = y_series
@@ -407,159 +421,12 @@ class WaterQualityModelingBatch:
         return X_train, X_test, y_train, y_test
 
     def spxy(self, data, label, test_size=0.2):
-        """
-        SPXY算法划分数据集（考虑X和Y空间的距离）
-        
-        Args:
-            data: shape (n_samples, n_features)
-            label: shape (n_samples, )
-            test_size: 测试集比例，默认: 0.2
-            
-        Returns:
-            X_train: (n_samples, n_features)
-            X_test: (n_samples, n_features)
-            y_train: (n_samples, )
-            y_test: (n_samples, )
-        """
-        # 确保 data 和 label 是 NumPy 数组
-        data = data.to_numpy() if isinstance(data, pd.DataFrame) else data
-        label = label.to_numpy() if isinstance(label, pd.Series) else label
-
-        # 备份原始数据和标签
-        x_backup = data
-        y_backup = label
-
-        M = data.shape[0]
-        N = round((1 - test_size) * M)
-        samples = np.arange(M)
-
-        # 归一化标签数据
-        label = (label - np.mean(label)) / np.std(label)
-        D = np.zeros((M, M))
-        Dy = np.zeros((M, M))
-
-        # 计算样本之间的距离
-        for i in range(M - 1):
-            xa = data[i, :]
-            ya = label[i]
-            for j in range((i + 1), M):
-                xb = data[j, :]
-                yb = label[j]
-                D[i, j] = np.linalg.norm(xa - xb)
-                Dy[i, j] = np.linalg.norm(ya - yb)
-
-        # 距离归一化
-        Dmax = np.max(D)
-        Dymax = np.max(Dy)
-        D = D / Dmax + Dy / Dymax
-
-        # 找到最远的两个点
-        maxD = D.max(axis=0)
-        index_row = D.argmax(axis=0)
-        index_column = maxD.argmax()
-
-        m = np.zeros(N, dtype=int)
-        m[0] = index_row[index_column]
-        m[1] = index_column
-
-        dminmax = np.zeros(N)
-        dminmax[1] = D[m[0], m[1]]
-
-        # 根据距离选择训练集
-        for i in range(2, N):
-            pool = np.delete(samples, m[:i])
-            dmin = np.zeros(M - i)
-            for j in range(M - i):
-                indexa = pool[j]
-                d = np.zeros(i)
-                for k in range(i):
-                    indexb = m[k]
-                    if indexa < indexb:
-                        d[k] = D[indexa, indexb]
-                    else:
-                        d[k] = D[indexb, indexa]
-                dmin[j] = np.min(d)
-            dminmax[i] = np.max(dmin)
-            index = np.argmax(dmin)
-            m[i] = pool[index]
-
-        m_complement = np.delete(samples, m)
-
-        # 划分训练集和测试集
-        X_train = data[m, :]
-        y_train = y_backup[m]
-        X_test = data[m_complement, :]
-        y_test = y_backup[m_complement]
-
-        return X_train, X_test, y_train, y_test
+        """SPXY算法划分数据集（委托至 src.core.utils.split_methods.spxy）"""
+        return spxy(data, label, test_size=test_size)
 
     def ks(self, data, label, test_size=0.2):
-        """
-        Kennard-Stone算法划分数据集
-        
-        Args:
-            data: shape (n_samples, n_features)
-            label: shape (n_sample, )
-            test_size: 测试集比例，默认: 0.2
-            
-        Returns:
-            X_train: (n_samples, n_features)
-            X_test: (n_samples, n_features)
-            y_train: (n_samples, )
-            y_test: (n_samples, )
-        """
-        # 确保 data 和 label 是 NumPy 数组
-        data = data.to_numpy() if isinstance(data, pd.DataFrame) else data
-        label = label.to_numpy() if isinstance(label, pd.Series) else label
-
-        M = data.shape[0]
-        N = round((1 - test_size) * M)
-        samples = np.arange(M)
-
-        D = np.zeros((M, M))
-
-        for i in range((M - 1)):
-            xa = data[i, :]
-            for j in range((i + 1), M):
-                xb = data[j, :]
-                D[i, j] = np.linalg.norm(xa - xb)
-
-        maxD = np.max(D, axis=0)
-        index_row = np.argmax(D, axis=0)
-        index_column = np.argmax(maxD)
-
-        m = np.zeros(N)
-        m[0] = np.array(index_row[index_column])
-        m[1] = np.array(index_column)
-        m = m.astype(int)
-        dminmax = np.zeros(N)
-        dminmax[1] = D[m[0], m[1]]
-
-        for i in range(2, N):
-            pool = np.delete(samples, m[:i])
-            dmin = np.zeros((M - i))
-            for j in range((M - i)):
-                indexa = pool[j]
-                d = np.zeros(i)
-                for k in range(i):
-                    indexb = m[k]
-                    if indexa < indexb:
-                        d[k] = D[indexa, indexb]
-                    else:
-                        d[k] = D[indexb, indexa]
-                dmin[j] = np.min(d)
-            dminmax[i] = np.max(dmin)
-            index = np.argmax(dmin)
-            m[i] = pool[index]
-
-        m_complement = np.delete(np.arange(data.shape[0]), m)
-
-        X_train = data[m, :]
-        y_train = label[m]
-        X_test = data[m_complement, :]
-        y_test = label[m_complement]
-
-        return X_train, X_test, y_train, y_test
+        """Kennard-Stone算法划分数据集（委托至 src.core.utils.split_methods.ks）"""
+        return ks(data, label, test_size=test_size)
 
     def split_data(self, X: np.ndarray, y: pd.Series, method: str = "random", 
                    test_size: float = 0.2, random_state: int = 42) -> Tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray]:
@@ -639,23 +506,22 @@ class WaterQualityModelingBatch:
         elif model_name == 'LightGBM':
             base_model.set_params(verbose=-1)
 
-        # 网格搜索 - 使用KFold代替StratifiedKFold
+        # 随机搜索 —— 替代穷举式 GridSearchCV，大幅降低寻优时间
         cv_strategy = KFold(n_splits=cv_folds, shuffle=True, random_state=random_state)
 
-        grid_search = GridSearchCV(
+        grid_search = RandomizedSearchCV(
             base_model,
             config['params'],
+            n_iter=10,
             cv=cv_strategy,
             scoring=scoring,
             n_jobs=safe_n_jobs,
-            verbose=1
+            random_state=random_state,
+            verbose=1,
         )
 
-        # 在训练集上训练模型
-        # with parallel_backend("threading", n_jobs=-1):
-        #     grid_search.fit(X_train, y_train)
         grid_search.fit(X_train, y_train)
-        
+
         # 获取最佳模型
         best_model = grid_search.best_estimator_
 

src/core/modeling/regression.py

@@ -315,7 +315,7 @@ def main():
     
     # 示例1: 使用所有回归方法分析光谱指数
     print("\n1. 光谱指数与叶绿素a的回归分析:")
-    sample_data = pd.read_csv(r"E:\code\WQ\pipeline_result\work_dir\5_training_spectra\water_quality_results.csv")
+    sample_data = pd.read_csv(r"E:\code\WQ\pipeline_result\work_dir\6_Spectral_Feature_Extraction\water_quality_results.csv")
     spectral_indices = ['Al10SABI','Am092Bsub']
     
     results1 = analyzer.batch_single_variable_regression(
@@ -323,7 +323,7 @@ def main():
         x_columns=spectral_indices,
         y_column='Chlorophyll',
         methods='all',
-        output_file=r'E:\code\WQ\pipeline_result\work_dir\5_training_spectra\spectral_indices_regression.csv'
+        output_file=r'E:\code\WQ\pipeline_result\work_dir\6_Spectral_Feature_Extraction\spectral_indices_regression.csv'
     )
     
     # # 示例2: 使用特定方法分析反射率波段
@@ -343,7 +343,7 @@ def main():
     best_models = analyzer.get_best_models_summary()
     if not best_models.empty:
         print(best_models[['x_variable', 'regression_method', 'r_squared', 'equation']].to_string(index=False))
-        best_models.to_csv(r'E:\code\WQ\pipeline_result\work_dir\5_training_spectra\best_models_summary.csv', index=False)
+        best_models.to_csv(r'E:\code\WQ\pipeline_result\work_dir\6_Spectral_Feature_Extraction\best_models_summary.csv', index=False)
         print("\n最佳模型汇总已保存到 'best_models_summary.csv'")
 #
 # def advanced_usage_example():

src/core/non_empirical_retrieval.py

@@ -246,8 +246,8 @@ def non_empirical_retrieval(algorithm, model_info_path, coor_spectral_path, outp
 
 if __name__ == "__main__":
     algorithm= "chl_a"
-    model_info_path= r"E:\code\WQ\pipeline_result\work_dir\5_training_spectra\8_non_empirical_models\SS\SS_chl_a.json"
-    coor_spectral_path= r"E:\code\WQ\pipeline_result\work_dir\10_sampling\sampling_spectra.csv"
+    model_info_path= r"E:\code\WQ\pipeline_result\work_dir\6_Spectral_Feature_Extraction\8_non_empirical_models\SS\SS_chl_a.json"
+    coor_spectral_path= r"E:\code\WQ\pipeline_result\work_dir\4_sampling\sampling_spectra.csv"
     output_path= r"E:\code\WQ\pipeline_result\work_dir\11_12_13_predictions\SS_chl_a.csv"
     wave_radius=5.0
     non_empirical_retrieval(algorithm, model_info_path, coor_spectral_path, output_path, wave_radius)

src/core/pipeline/__init__.py

@@ -0,0 +1,24 @@
+# -*- coding: utf-8 -*-
+"""
+Pipeline 调度核心：基于 Context 的内存级依赖注入。
+
+设计目标：
+  - 用 PipelineContext 替代 dict 散落传参（9 步主路径 + 14 个 step 共享同一份 ctx）
+  - 14 个 step 声明式描述（StepSpec），便于 Web / 异步 / 单元测试复用
+  - 不绑定具体 Pipeline 实现（duck-typed），WorkerThread / Web API / 单测可共用
+"""
+
+from .context import (
+    PipelineContext,
+    STEP_MAP_OLD_TO_NEW, STEP_MAP_NEW_TO_OLD,
+    resolve_step_id, ALL_STEP_IDS,
+)
+from .runner import (
+    StepSpec, PIPELINE_STEPS, PipelineRunner, PipelineHalt,
+)
+
+__all__ = [
+    "PipelineContext", "StepSpec", "PIPELINE_STEPS", "PipelineRunner", "PipelineHalt",
+    "STEP_MAP_OLD_TO_NEW", "STEP_MAP_NEW_TO_OLD",
+    "resolve_step_id", "ALL_STEP_IDS",
+]

src/core/pipeline/context.py

@@ -0,0 +1,148 @@
+# -*- coding: utf-8 -*-
+"""
+PipelineContext：内存级数据载体，跨 14 个 step 传递路径与元信息。
+
+设计原则：
+  - 所有路径字段以 `_path` 为后缀（与 step 方法形参命名约定一致）
+  - 字段值可缺省（None），由 StepSpec.requires 在调度时注入
+  - dataclass + field(default_factory=dict) 支持原地增删
+  - 不放 GUI 状态（避免循环依赖）
+  - 不绑具体 step 方法（duck-typed cancellation / log append）
+"""
+
+from __future__ import annotations
+from dataclasses import dataclass, field
+from typing import Any, Dict, List, Optional, Set
+
+
+# ============================================================
+# 步骤命名映射（定义在叶子节点，打破循环依赖）
+# ============================================================
+
+STEP_MAP_OLD_TO_NEW: Dict[str, str] = {
+    "step5_5":   "step7",
+    "step6_5":   "step8_non_empirical_modeling",
+    "step6_75":  "step9",
+    "step8_5":   "step11",
+    "step7":     "step8",
+    "step8":     "step7",
+    "step9":     "step14",
+    "step10":    "step4",
+    "step11_ml": "step10",
+    "step11":    "step11",
+}
+
+STEP_MAP_NEW_TO_OLD: Dict[str, str] = {v: k for k, v in STEP_MAP_OLD_TO_NEW.items()}
+
+ALL_STEP_IDS: Set[str] = set(STEP_MAP_OLD_TO_NEW.keys()) | set(STEP_MAP_OLD_TO_NEW.values())
+
+
+def resolve_step_id(step_id: str) -> str:
+    """将任意 step_id 转换为标准新格式。"""
+    if step_id in STEP_MAP_OLD_TO_NEW:
+        return STEP_MAP_OLD_TO_NEW[step_id]
+    return step_id
+
+
+@dataclass
+class PipelineContext:
+    """流水线运行上下文（在 14 个 step 之间传递的内存字典）
+
+    字段命名约定：
+      - 路径类字段名 = panel key 名 = step 形参名（全链路无翻译）
+      - 训练/产物 CSV 用 `_path` 后缀（如 training_csv_path / water_mask_path）
+      - 入参影像/CSV 沿用 panel 原名（img_path / csv_path），无 `_path` 后缀
+      - 目录类字段无 `_path` 后缀（如 models_dir / prediction_dir）
+      - 元信息字段无后缀（如 user_config / status / log）
+    """
+
+    # ── 11 个 step 的入参/产物（按 step 顺序排列；字段名 = panel key = step 形参） ──
+    img_path: Optional[str] = None                # Step 1/2/3 入参：原始影像
+    water_mask_path: Optional[str] = None         # Step 1 出 → Step 2/3/7 入
+    glint_mask_path: Optional[str] = None         # Step 2 出 → Step 3/7 入
+    deglint_img_path: Optional[str] = None        # Step 3 出 → Step 5/7 入
+    csv_path: Optional[str] = None                # Step 4/5/6_5/6_75 入参：原始/训练 CSV
+    processed_csv_path: Optional[str] = None      # Step 4 出 → Step 5 入
+    training_csv_path: Optional[str] = None       # Step 5 出 → Step 5_5/6/6_5/6_75 入
+    boundary_path: Optional[str] = None           # Step 5 入参：边界 SHP（panel step5 名）
+    indices_path: Optional[str] = None            # Step 5.5 出
+    sampling_csv_path: Optional[str] = None       # Step 7 出 → Step 8/8_5/8_75/9 入
+    prediction_csv_path: Optional[str] = None     # Step 8 出 → Step 9 入
+    distribution_map_path: Optional[str] = None   # Step 9 出
+    boundary_shp_path: Optional[str] = None       # Step 9 入参：边界 SHP（panel step9 名）
+    formula_csv_path: Optional[str] = None        # Step 8_75 入参：公式 CSV
+
+    # ── 目录类（命名不带 _path 以示区别） ──
+    models_dir: Optional[str] = None
+    prediction_dir: Optional[str] = None
+    work_dir: Optional[str] = None
+
+    # ── Step 6 训练产物（AutoML 模式有，常规模式为空） ──
+    model_files: List[str] = field(default_factory=list)
+
+    # ── 元信息（三件套：用户传的配置 / 取消事件 / 状态） ──
+    user_config: Dict[str, Any] = field(default_factory=dict)
+    cancel_event: Optional[Any] = None            # duck-typed threading.Event / asyncio.Event
+    status: Dict[str, str] = field(default_factory=dict)   # {step_id: 'start'/'completed'/'skipped'/'error'}
+    log: List[str] = field(default_factory=list)
+
+    # ── 诊断 ──
+    step_timings: Dict[str, float] = field(default_factory=dict)
+    pipeline_start_time: Optional[float] = None
+    pipeline_end_time: Optional[float] = None
+    last_error: Optional[str] = None
+
+    # ── 错误汇总（全流程结束后可用） ──
+    error_summary: List[tuple[str, str]] = field(default_factory=list)
+    # ── 出错时立即停止全流程（默认 False：继续后续步骤） ──
+    breakpoint_on_error: bool = False
+    # ── ★ 智能补全锁定步骤列表（由 _auto_fill_missing_steps 自动开启的步骤） ──
+    # GUI 层读取此字段，在运行期间禁用对应面板的启用复选框
+    locked_steps: List[str] = field(default_factory=list)
+
+    # ============================================================
+    # 读写辅助
+    # ============================================================
+
+    def step_id(self, step_id: str) -> str:
+        """将任意 step_id（可能是旧名）转换为标准新格式。
+
+        用法示例：
+            ctx.status[ctx.step_id('step6_5')]   # 'step8_non_empirical_modeling'
+            ctx.user_config[ctx.step_id('step8_5')]  # 'step11'
+        """
+        if step_id in STEP_MAP_OLD_TO_NEW:
+            return STEP_MAP_OLD_TO_NEW[step_id]
+        return step_id
+
+    def set(self, key: str, value: Any) -> None:
+        """原地写入任意属性。
+
+        允许动态字段（如 'report_path'）直接挂在 __dict__ 上，
+        避免因静态字段缺失而抛 AttributeError。
+        """
+        object.__setattr__(self, key, value)
+
+    def get(self, key: str, default: Any = None) -> Any:
+        """原地读出，缺 key 不抛错。"""
+        return getattr(self, key, default)
+
+    def is_cancelled(self) -> bool:
+        """统一软取消检查入口（duck-typed）。
+
+        支持：
+          - threading.Event（.is_set()）
+          - asyncio.Event（loop-bound，is_set 同步接口存在）
+          - 自定义 .is_set() / .cancelled 属性
+        """
+        ev = self.cancel_event
+        if ev is None:
+            return False
+        is_set = getattr(ev, "is_set", None)
+        if callable(is_set):
+            return bool(is_set())
+        return bool(getattr(ev, "cancelled", False))
+
+    def append_log(self, msg: str) -> None:
+        """写入日志列表（也用于主进程 stdout 调试）。"""
+        self.log.append(msg)

src/core/pipeline/runner.py

@@ -0,0 +1,650 @@
+# -*- coding: utf-8 -*-
+"""
+PipelineRunner：基于 StepSpec 声明式调度 14 个 step。
+
+设计要点：
+  - StepSpec 声明 requires（ctx 字段名列表）+ produces（ctx 字段名列表）
+  - 命名约定：ctx 字段名 == panel key 名 == step 形参名（全链路无翻译）
+  - 步骤命名：step_id 格式为 stepN 或 stepN_suffix（无小数位），method_name 与 step_id 对齐
+  - 调度顺序：按 PIPELINE_STEPS 列表顺序，requires 缺则 skip
+  - 软取消：在每个 step 前检查 ctx.is_cancelled()
+  - 断点续跑：spec.output_file 已落盘则跳过执行
+  - 错误汇总：全流程结束后 error_summary 记录所有 step 的异常
+  - 预检：run() 入口硬校验 step1 img_path；其余依赖通过智能补全 + 软警告处理
+  - PipelineHalt：外层 run() 不 catch，触发循环 break，实现硬终止
+  - STEP_MAP：旧 step_id → 新 step_id 双向映射，供 GUI 配置兼容使用
+  - duck-typed pipeline：runner 只调 getattr(pipeline, method_name)，不强依赖类层级
+"""
+
+from __future__ import annotations
+import inspect
+import logging
+import os
+import time
+from dataclasses import dataclass, field
+from typing import Any, Dict, List, Optional, Sequence
+
+from .context import PipelineContext, STEP_MAP_OLD_TO_NEW, STEP_MAP_NEW_TO_OLD, resolve_step_id
+
+logger = logging.getLogger(__name__)
+
+
+# ============================================================
+# 终止异常（外层 run() 不 catch，触发循环 break）
+# ============================================================
+
+class PipelineHalt(Exception):
+    """不可恢复的错误，在 run() 循环中抛出后直接 break，不走 Exception 处理分支。
+
+    适用场景：
+      - GUI 层通过 _notify 弹窗拦截后主动抛出的硬终止信号
+    """
+    pass
+
+
+# ============================================================
+# StepSpec 声明式描述
+# ============================================================
+
+@dataclass
+class StepSpec:
+    """单个 step 的元信息（声明式，避免硬编码）"""
+    step_id: str
+    method_name: str
+    requires: List[str]                                  # PipelineContext 字段名列表
+    produces: List[str] = field(default_factory=list)    # 写入 ctx 的字段名列表
+    enabled: bool = True
+    parameter_map: Dict[str, str] = field(default_factory=dict)
+    # 当 requires 中任一字段为 None 时是否跳过；默认 True（缺输入就 skip）
+    skip_when_missing: bool = True
+    # 备注（仅用于文档生成 / 调试输出）
+    description: str = ""
+    # ★ 断点续跑：产物文件路径，支持 {work_dir} 占位符（运行时解析）
+    output_file: Optional[str] = None
+    # ★ 预检用：需要验证磁盘文件实际存在的 ctx key 列表
+    required_input_files: List[str] = field(default_factory=list)
+
+
+# ============================================================
+# 14 个 step 的声明表（顺序即调度顺序）
+# step_id / method_name 均不含小数位，与前端显示对齐
+# output_file / required_input_files 使用 {work_dir} 占位符，由 _resolve_path 展开
+# ============================================================
+
+PIPELINE_STEPS: List[StepSpec] = [
+    StepSpec(
+        step_id="step1", method_name="step1_generate_water_mask",
+        requires=["img_path"], produces=["water_mask_path"],
+        required_input_files=["img_path"],
+        output_file="{work_dir}/1_water_mask/water_mask.dat",
+        description="水域掩膜生成（NDWI 或 SHP）",
+    ),
+    StepSpec(
+        step_id="step2", method_name="step2_find_glint_area",
+        requires=["img_path", "water_mask_path"], produces=["glint_mask_path"],
+        required_input_files=["img_path", "water_mask_path"],
+        output_file="{work_dir}/2_Glint_Detection/severe_glint_area.dat",
+        description="耀斑区域检测",
+    ),
+    StepSpec(
+        step_id="step3", method_name="step3_remove_glint",
+        requires=["img_path", "water_mask_path", "glint_mask_path"],
+        produces=["deglint_img_path"],
+        required_input_files=["img_path", "water_mask_path", "glint_mask_path"],
+        output_file="{work_dir}/3_deglint/deglint.bsq",
+        description="耀斑去除",
+    ),
+    StepSpec(
+        step_id="step4", method_name="step5_process_csv",
+        requires=["csv_path"], produces=["processed_csv_path"],
+        required_input_files=["csv_path"],
+        output_file="{work_dir}/5_Data_Cleaning/processed_data.csv",
+        description="CSV 异常值清洗",
+    ),
+    StepSpec(
+        step_id="step5", method_name="step6_extract_spectra",
+        requires=["deglint_img_path", "processed_csv_path", "csv_path", "boundary_path", "glint_mask_path"],
+        produces=["training_csv_path"],
+        parameter_map={
+            "processed_csv_path": "csv_path",
+            "csv_path": "_raw_csv_ignored",
+        },
+        skip_when_missing=False,
+        required_input_files=["deglint_img_path", "processed_csv_path", "boundary_path", "glint_mask_path"],
+        output_file="{work_dir}/6_Spectral_Feature_Extraction/training_spectra.csv",
+        description="实测样本点光谱提取",
+    ),
+    StepSpec(
+        step_id="step7", method_name="step7_calc_indices",
+        requires=["training_csv_path"], produces=["indices_path", "trad_indices_dir"],
+        required_input_files=["training_csv_path"],
+        output_file="{work_dir}/7_Water_Quality_Indices/training_spectra_indices.csv",
+        description="水质参数指数计算（双轨输出：A轨宽表 + B轨单文件）",
+    ),
+    StepSpec(
+        step_id="step8", method_name="step8_train_ml",
+        requires=["training_csv_path"], produces=["models_dir"],
+        required_input_files=["training_csv_path"],
+        output_file="{work_dir}/8_Supervised_Model_Training/best_models.pkl",
+        description="ML 建模（GridSearchCV / AutoML）",
+    ),
+    StepSpec(
+        step_id="step8_non_empirical_modeling",
+        method_name="step8_non_empirical_modeling",
+        requires=["training_csv_path"], produces=["models_dir"],
+        parameter_map={"training_csv_path": "csv_path"},
+        required_input_files=["training_csv_path"],
+        output_file="{work_dir}/8_Non_Empirical_Regression/non_empirical_models.pkl",
+        description="非经验统计回归",
+    ),
+    StepSpec(
+        step_id="step9", method_name="step9_watercolor_inversion",
+        requires=["deglint_img_path", "water_mask_path"], produces=["watercolor_index_dir"],
+        required_input_files=["deglint_img_path"],
+        output_file="{work_dir}/9_WaterColor_Index_Images",
+        description="水色指数反演（BSQ 影像直接处理）",
+    ),
+    StepSpec(
+        step_id="step10", method_name="step4_sampling",
+        requires=["deglint_img_path", "water_mask_path"], produces=["sampling_csv_path"],
+        required_input_files=["deglint_img_path", "water_mask_path"],
+        output_file="{work_dir}/4_sampling/sampling_spectra.csv",
+        description="整景密集采样点生成 + 光谱提取",
+    ),
+    StepSpec(
+        step_id="step11_ml", method_name="step9_predict_ml",
+        requires=["sampling_csv_path", "models_dir"], produces=["prediction_csv_path"],
+        required_input_files=["sampling_csv_path", "models_dir"],
+        output_file="{work_dir}/11_12_13_predictions/prediction_results.csv",
+        description="ML 模型预测（采样点）",
+    ),
+    StepSpec(
+        step_id="step11", method_name="step11_non_empirical_prediction",
+        requires=["sampling_csv_path", "models_dir"], produces=["prediction_dir"],
+        parameter_map={"models_dir": "non_empirical_models_dir"},
+        required_input_files=["sampling_csv_path", "models_dir"],
+        output_file="{work_dir}/11_12_13_predictions/non_empirical_predictions",
+        description="非经验模型预测",
+    ),
+    StepSpec(
+        step_id="step14", method_name="step10_map",
+        requires=["prediction_csv_path", "boundary_shp_path"],
+        produces=["distribution_map_path"],
+        required_input_files=["prediction_csv_path", "boundary_shp_path"],
+        output_file="{work_dir}/distribution_map.png",
+        description="克里金插值成图",
+    ),
+]
+
+
+# ============================================================
+# PipelineRunner：执行者
+# ============================================================
+
+class PipelineRunner:
+    """按 StepSpec 调度 14 个 step 方法，支持软取消 + 断点续跑 + 错误汇总。
+
+    用法：
+        ctx = PipelineContext(img_path=..., work_dir=..., user_config=config)
+        runner = PipelineRunner(pipeline_instance)
+        result_ctx = runner.run(ctx, config=config)  # 预检通过后开始执行
+        print(result_ctx.error_summary)           # [(step_id, error_msg), ...]
+    """
+
+    def __init__(self, pipeline, steps: Optional[Sequence[StepSpec]] = None):
+        self.pipeline = pipeline
+        self.steps: List[StepSpec] = list(steps) if steps else list(PIPELINE_STEPS)
+
+    # ------------------------------------------------------------------
+    # 主入口
+    # ------------------------------------------------------------------
+
+    def run(self, ctx: PipelineContext, config=None, skip_list: Optional[List[str]] = None) -> PipelineContext:
+        self.config = config or {}
+        skip_list = skip_list or []
+        logger.info("开始运行完整流程 (Runner 调度模式)...")
+
+        ctx.pipeline_start_time = time.time()
+        error_summary: List[tuple[str, str]] = []
+        skip_set = set(skip_list) if skip_list else set()
+
+        # ── ★ Step1 img_path 硬校验（缺失则立即终止整个流程） ──
+        if not ctx.get("img_path"):
+            msg = "【全流程预检失败】缺少参考影像路径 (img_path)，流程无法启动。"
+            ctx.append_log(f"[RUNNER] {msg}")
+            self._notify_step("全流程", "error", msg)
+            ctx.last_error = msg
+            ctx.pipeline_end_time = time.time()
+            return ctx
+
+        # ── ★ 智能补全：扫描 work_dir 默认产物路径，回填 ctx ──
+        self._scan_workdir_outputs(ctx)
+
+        # ── ★ 自动补全缺失步骤：work_dir 有产物则强制开启 + 回填路径 ──
+        self._auto_fill_missing_steps(ctx)
+
+        # ── 软预检警告（不再阻断，仅记录日志）──
+        self._preflight_warnings(ctx)
+
+        # 断点续跑预扫描：ctx 已有产物则记录诊断日志
+        self._restore_outputs_from_ctx(ctx)
+
+        # 1. 暴力上下文注入：将 GUI config 中的所有参数强行塞入 ctx，防丢失
+        for step_id, cfg in self.config.items():
+            if isinstance(cfg, dict):
+                for k, v in cfg.items():
+                    if k != 'enabled' and v:
+                        setattr(ctx, k, v)
+
+        # 2. 构建依赖提供者映射 (Provider Map)
+        provider_map = {}
+        for step in self.steps:
+            for prod in step.produces:
+                provider_map[prod] = step
+
+        # 3. 强力依赖级联唤醒 (Auto-Wakeup Engine)
+        changed = True
+        woke_up_steps = []
+        while changed:
+            changed = False
+            for step in self.steps:
+                if step.step_id in skip_set:
+                    continue  # 用户强踢的，绝不唤醒
+
+                step_cfg = self.config.setdefault(step.step_id, {})
+                if not step_cfg.get('enabled', True):
+                    continue
+
+                for req in step.requires:
+                    # 如果上下文缺这个参数
+                    if not (hasattr(ctx, req) and getattr(ctx, req)):
+                        provider = provider_map.get(req)
+                        if provider and provider.step_id not in skip_set:
+                            prov_cfg = self.config.setdefault(provider.step_id, {})
+                            if not prov_cfg.get('enabled', True):
+                                prov_cfg['enabled'] = True
+                                changed = True
+                                woke_up_steps.append(provider.step_id)
+                                logger.info(f"[*] 自动唤醒: {provider.step_id} (为下游提供 {req})")
+
+        if woke_up_steps:
+            logger.info(f"★ 依赖唤醒完成，共唤醒 {len(woke_up_steps)} 个次/步骤")
+
+        # 4. 正式执行流水线
+        for step in self.steps:
+            # ── 软取消 ──
+            if ctx.is_cancelled():
+                ctx.append_log(f"[RUNNER] 收到取消信号，提前终止 @ {step.step_id}")
+                break
+
+            if step.step_id in skip_set:
+                ctx.status[step.step_id] = "user_skipped"
+                ctx.append_log(
+                    f"\n{'='*60}\n"
+                    f"  ⚠ 用户强制跳过: {step.step_id}（{step.description}）\n"
+                    f"  原因：用户在预检弹窗中勾选「忽略」，已确认跳过\n"
+                    f"{'='*60}\n"
+                )
+                self._notify_step(step.step_id, "skipped", "用户强制跳过（预检弹窗）")
+                continue
+
+            step_cfg = self.config.get(step.step_id, {})
+            if not step_cfg.get('enabled', True):
+                continue
+
+            # 4.1 检查磁盘产物：如果已落盘，恢复上下文并跳过（拒绝静默跳过，必须打日志）
+            if step.output_file and os.path.exists(step.output_file):
+                for prod in step.produces:
+                    if not (hasattr(ctx, prod) and getattr(ctx, prod)):
+                        setattr(ctx, prod, step.output_file)
+                ctx.status[step.step_id] = "skipped"
+                ctx.append_log(f"[CACHE] 产物已存在，跳过运行并恢复上下文: {step.step_id}")
+                self._notify_step(step.step_id, "skipped", "产物已存在（断点续跑）")
+                continue
+
+            # 4.2 依赖死线检查
+            missing = [req for req in step.requires if not (hasattr(ctx, req) and getattr(ctx, req))]
+            if missing:
+                ctx.status[step.step_id] = "skipped"
+                reason = f"缺少必要的上下文参数，自动跳过: {missing}"
+                ctx.append_log(f"[RUNNER] 跳过 {step.step_id}，仍缺少必要参数: {missing}")
+                self._notify_step(step.step_id, "skipped", reason)
+                continue
+
+            # 4.3 真正执行
+            ctx.append_log(f"[START] 正在执行步骤: {step.step_id}")
+            self._notify_step(step.step_id, "running", f"正在执行: {step.description}")
+            try:
+                method = getattr(self.pipeline, step.method_name)
+
+                sig = inspect.signature(method)
+                kwargs = {}
+                current_step_cfg = self.config.get(step.step_id, {})
+
+                for param_name in sig.parameters:
+                    # 优先级 1：直接使用当前步骤专属配置中的值
+                    if param_name in current_step_cfg:
+                        kwargs[param_name] = current_step_cfg[param_name]
+                        continue
+                        
+                    # 优先级 1.5：【核心修复】硬隔离 output_file，防止被其他步骤的同名变量污染
+                    if param_name == 'output_file' and hasattr(step, 'output_file') and step.output_file:
+                        work_dir = getattr(ctx, 'work_dir', '')
+                        kwargs[param_name] = step.output_file.format(work_dir=work_dir)
+                        continue
+                        
+                    # 优先级 2：处理跨步骤的映射逻辑
+                    ctx_key = param_name
+                    if hasattr(step, 'parameter_map') and step.parameter_map:
+                        for k, v in step.parameter_map.items():
+                            if v == param_name:
+                                ctx_key = k
+                                break
+                    # 优先级 3：从全局大背包 ctx 中取（排在最后）
+                    if hasattr(ctx, ctx_key):
+                        kwargs[param_name] = getattr(ctx, ctx_key)
+
+                # 使用解包后的关键字参数调用底层函数
+                result = method(**kwargs)
+
+                # 【产物接力 1】：如果底层函数返回了字典，直接合并到上下文
+                if isinstance(result, dict):
+                    for k, v in result.items():
+                        setattr(ctx, k, v)
+
+                # 【产物接力 2】：强制通过 StepSpec 的 output_file 模板注入
+                if hasattr(step, 'output_file') and step.output_file:
+                    work_dir = getattr(ctx, 'work_dir', '')
+                    actual_out_path = step.output_file.format(work_dir=work_dir)
+                    for prod in step.produces:
+                        if not hasattr(ctx, prod) or not getattr(ctx, prod):
+                            setattr(ctx, prod, actual_out_path)
+                            logger.info(f"[产物接力] 登记 {prod} = {actual_out_path}")
+            except PipelineHalt:
+                ctx.status[step.step_id] = "error"
+                ctx.append_log(f"[RUNNER] PipelineHalt 硬终止 @ {step.step_id}")
+                self._notify_step(step.step_id, "error", "预检失败，硬终止")
+                break
+            except Exception as e:
+                ctx.status[step.step_id] = "error"
+                error_summary.append((step.step_id, str(e)))
+                ctx.last_error = f"{step.step_id}: {e!r}"
+                ctx.append_log(f"[ERROR] 步骤 {step.step_id} 执行崩溃: {str(e)}")
+                self._notify_step(step.step_id, "error", str(e))
+                break
+
+        ctx.pipeline_end_time = time.time()
+        ctx.error_summary = error_summary
+        return ctx
+
+    # ------------------------------------------------------------------
+    # ★ 智能补全：工作目录产物扫描
+    # ------------------------------------------------------------------
+
+    def _scan_workdir_outputs(self, ctx: PipelineContext) -> None:
+        """扫描 work_dir 下所有步骤的默认产物路径，若存在则回填 ctx。
+
+        利用 spec.output_file 的 {work_dir} 占位符，展开为实际绝对路径。
+        存在则写入对应的 ctx 字段（produces），供后续步骤直接使用。
+        已在 ctx 中有值的字段不会被覆盖。
+        """
+        work_dir = ctx.get("work_dir") or ""
+        if not work_dir:
+            return
+
+        for spec in self.steps:
+            if not spec.produces:
+                continue
+            for produce_key in spec.produces:
+                if ctx.get(produce_key):
+                    continue  # 已有人工填写的值，不覆盖
+                resolved = self._resolve_path(spec.output_file, ctx)
+                if resolved and os.path.exists(resolved):
+                    ctx.set(produce_key, resolved)
+                    ctx.append_log(
+                        f"[AUTO_FILL] 检测到已有产物，回填 {produce_key} = {resolved}"
+                    )
+
+    # ------------------------------------------------------------------
+    # ★ 智能补全：强制开启被静默跳过的步骤
+    # ------------------------------------------------------------------
+
+    def _auto_fill_missing_steps(self, ctx: PipelineContext) -> None:
+        """检查所有 disabled 步骤。
+
+        若某步骤的 output_file 已在 work_dir 落盘（断点续跑），
+        说明该步骤之前已完成但被用户在 GUI 中禁用了。
+        此时系统自动重开启该步骤（forced=True），并将其加入 locked_steps。
+
+        同时，将已落盘的产物路径回填到对应的 ctx 字段，
+        确保下游步骤能正常拿到输入。
+
+        阻断性缺失（step1 img_path）已在 run() 入口硬校验，此处不处理。
+        """
+        newly_locked: List[str] = []
+
+        for spec in self.steps:
+            if spec.enabled:
+                continue  # 用户主动开启的步骤不受影响
+            skip_set = getattr(ctx, '_skip_set', set())
+            if spec.step_id in skip_set:
+                continue  # 用户在 PreflightDialog 中手动忽略的步骤不自动补全
+
+            resolved = self._resolve_path(spec.output_file, ctx)
+            if resolved and os.path.exists(resolved):
+                # ── 该步骤已有产物但被禁用 → 自动开启 ──
+                spec.enabled = True
+                ctx.locked_steps.append(spec.step_id)
+                newly_locked.append(spec.step_id)
+
+                # 回填所有产物字段到 ctx
+                for produce_key in spec.produces:
+                    if not ctx.get(produce_key):
+                        ctx.set(produce_key, resolved)
+                        ctx.append_log(
+                            f"[AUTO_FILL] 强制开启并回填 {spec.step_id} 产物 {produce_key} = {resolved}"
+                        )
+
+                ctx.append_log(
+                    f"\n{'='*60}\n"
+                    f"  ⚡ 智能补全：步骤 {spec.step_id}（{spec.description}）\n"
+                    f"  原因：该步骤在 work_dir 中已有产物但被您在 GUI 中禁用了。\n"
+                    f"  操作：系统已自动开启该步骤，产物路径已回填。\n"
+                    f"  注意：运行期间该步骤已被锁定，您无法临时关闭。\n"
+                    f"{'='*60}\n"
+                )
+
+        if newly_locked:
+            self._notify_step(
+                "全流程",
+                "info",
+                f"智能补全已自动开启 {len(newly_locked)} 个步骤：{newly_locked}"
+            )
+
+    def _resolve_output_for_key(
+        self, produce_key: str, ctx: PipelineContext
+    ) -> Optional[str]:
+        """根据 produces key 查找对应步骤的 output_file 并展开路径。"""
+        for spec in self.steps:
+            if produce_key in spec.produces:
+                return self._resolve_path(spec.output_file, ctx)
+        return None
+
+    def _scan_single_step_outputs(
+        self, spec: StepSpec, ctx: PipelineContext
+    ) -> None:
+        """扫描单个步骤的 work_dir 产物，回填 ctx（不覆盖已有值）。"""
+        if not spec.produces:
+            return
+        for produce_key in spec.produces:
+            if ctx.get(produce_key):
+                continue
+            resolved = self._resolve_path(spec.output_file, ctx)
+            if resolved and os.path.exists(resolved):
+                ctx.set(produce_key, resolved)
+                ctx.append_log(
+                    f"[AUTO_FILL] 依赖唤醒后检测到产物，回填 {produce_key} = {resolved}"
+                )
+
+    # ------------------------------------------------------------------
+    # 软预检警告（不再阻断）
+    # ------------------------------------------------------------------
+
+    def _preflight_warnings(self, ctx: PipelineContext) -> None:
+        """软预检警告：遍历所有步骤，检测可预见的运行时跳过。
+
+        所有缺失均以 warning 记录日志，不抛异常，不阻止执行。
+        GUI 层可通过回调函数 _notify_step 向用户展示警告列表。
+        """
+        warnings: List[str] = []
+
+        for spec in self.steps:
+            if not spec.enabled:
+                continue
+
+            # ── Step4 csv_path 缺失警告 ──
+            if spec.step_id == "step4":
+                if not ctx.get("csv_path"):
+                    warnings.append(
+                        f"[{spec.step_id}] 缺少实测水质数据 (csv_path)，"
+                        "步骤 5-9 将被自动跳过"
+                    )
+
+            # ── 磁盘文件缺失警告（已填充 ctx 但文件实际不存在）──
+            for ctx_key in spec.required_input_files:
+                value = ctx.get(ctx_key)
+                if not value:
+                    continue
+                if not os.path.exists(value):
+                    warnings.append(
+                        f"[{spec.step_id}] 磁盘文件缺失（但 ctx 已回填）: {ctx_key} = {value}"
+                    )
+
+        if warnings:
+            detail = "\n".join(f"  - {w}" for w in warnings)
+            ctx.append_log(
+                f"[RUNNER] 【软预检警告】（流程将继续执行，缺失项将被自动跳过）\n{detail}"
+            )
+            self._notify_step("全流程", "warning", f"预检警告：{len(warnings)} 项\n{detail}")
+
+    # ------------------------------------------------------------------
+    # 单步调用
+    # ------------------------------------------------------------------
+
+    def _invoke(self, spec: StepSpec, ctx: PipelineContext) -> None:
+        """调一个 step 方法：ctx 路径 → 形参；产出 → ctx 字段。"""
+        ctx.append_log(
+            f"[DEBUG] Step {spec.step_id} requires: {spec.requires}, "
+            f"actual ctx data: {[ctx.get(k) for k in spec.requires]}"
+        )
+        method = getattr(self.pipeline, spec.method_name, None)
+        if method is None:
+            ctx.append_log(f"[RUNNER] 步骤方法缺失: {spec.method_name}（跳过）")
+            ctx.status[spec.step_id] = "skipped"
+            return
+
+        # 1) 把 ctx 路径作为形参注入
+        kwargs: Dict[str, Any] = {}
+        for ctx_key in spec.requires:
+            param_name = spec.parameter_map.get(ctx_key, self._default_param_name(ctx_key))
+            kwargs[param_name] = ctx.get(ctx_key)
+
+        # 2) 允许用户在 ctx.user_config[step_id] 覆盖/补充（非空值才覆盖）
+        user_overrides = ctx.user_config.get(spec.step_id) or {}
+        if isinstance(user_overrides, dict):
+            for k, v in user_overrides.items():
+                if v is not None and v != "":
+                    kwargs[k] = v
+
+        # 3) 状态置 start
+        ctx.append_log(
+            f"[RUNNER] -> {spec.method_name}({list(kwargs.keys())})"
+        )
+        ctx.status[spec.step_id] = "start"
+        self._notify_step(spec.step_id, "start", spec.method_name)
+
+        # 4) 执行（外层 run() 统一捕获异常）
+        t0 = time.time()
+        result = method(**kwargs)
+        ctx.status[spec.step_id] = "completed"
+        ctx.step_timings[spec.step_id] = time.time() - t0
+
+        # 5) 产出收割
+        self._harvest(spec, result, ctx)
+        self._notify_step(
+            spec.step_id, "completed",
+            str(result)[:200] if result is not None else "",
+        )
+
+    # ------------------------------------------------------------------
+    # 产出收割
+    # ------------------------------------------------------------------
+
+    def _harvest(self, spec: StepSpec, result: Any, ctx: PipelineContext) -> None:
+        """把 step 方法返回值灌入 ctx 的 produces 字段。"""
+        if not spec.produces:
+            return
+        if isinstance(result, dict):
+            for produce_key in spec.produces:
+                if produce_key in result:
+                    ctx.set(produce_key, result[produce_key])
+        elif result is not None:
+            ctx.set(spec.produces[0], result)
+
+    # ------------------------------------------------------------------
+    # 断点续跑辅助
+    # ------------------------------------------------------------------
+
+    def _resolve_path(
+        self, template: Optional[str], ctx: PipelineContext
+    ) -> Optional[str]:
+        """解析模板中的 {work_dir} 占位符，返回展开后的绝对路径或 None。"""
+        if not template:
+            return None
+        work_dir = ctx.get("work_dir") or ""
+        try:
+            return template.format(work_dir=work_dir)
+        except (KeyError, ValueError):
+            return template
+
+    def _restore_outputs_from_ctx(self, ctx: PipelineContext) -> None:
+        """诊断日志：记录 ctx 中已有的非 None 产物。"""
+        for spec in self.steps:
+            if not (spec.enabled and spec.produces):
+                continue
+            for key in spec.produces:
+                val = ctx.get(key)
+                if val:
+                    ctx.append_log(
+                        f"[RUNNER] 断点续跑检测: {spec.step_id} 已有 {key} = {val}"
+                    )
+
+    def _restore_ctx_from_output(
+        self, spec: StepSpec, resolved_path: str, ctx: PipelineContext
+    ) -> None:
+        """断点跳过时：将已存在的 output_file 写回 ctx 所有 produces 字段，供下游使用。
+
+        接力棒断链修复：遍历 spec.produces 逐一注册，不遗漏任何下游可能依赖的 key。
+        """
+        if not spec.produces:
+            return
+        for produce_key in spec.produces:
+            ctx.set(produce_key, resolved_path)
+
+    # ------------------------------------------------------------------
+    # 工具
+    # ------------------------------------------------------------------
+
+    @staticmethod
+    def _default_param_name(ctx_key: str) -> str:
+        """默认原样返回 ctx 键名作为形参名。特殊缩写由 parameter_map 显式处理。"""
+        return ctx_key
+
+    def _notify_step(self, step_id: str, status: str, message: str) -> None:
+        """通过 pipeline.callback 通知 GUI 当前步骤状态。"""
+        notify = getattr(self.pipeline, "_notify", None)
+        if callable(notify):
+            try:
+                notify(step_id, status, message)
+            except Exception:
+                pass

src/core/prediction/automl_trainer.py

@@ -0,0 +1,544 @@
+# -*- coding: utf-8 -*-
+"""
+Optuna + 智能子采样 AutoML 训练器（路线 B 防爆引擎）。
+
+为什么需要这个：
+  - 老路径：11 预处理 × 4 模型 × 3 划分 = 132 组 GridSearchCV
+    对中小数据集 10 分钟+，对大数据集 5w+ 行 直接 OOM
+  - AutoML 路径：1 预处理 × N 模型（Optuna 调超参），用智能子采样避开 OOM
+    再用最优超参在**全量数据**上 refit，最终保存单一模型
+
+设计要点：
+  - 入口 train_with_automl(csv, feature_start_column, model_names, ...)
+  - AutoMLResult dataclass 返回（每个目标列一份）
+  - smart_subsample：N > max_samples 时随机下采样
+  - 失败兜底：optuna 未装 / 全 trial 失败 → fallback 到 WaterQualityModelingBatch
+  - 文件命名规范：{target}_{preprocess}_{model}_AUTOML.joblib
+  - save_data["metadata"]["automl"] = True 标记
+
+调用：
+    from src.core.prediction.automl_trainer import train_with_automl
+    results = train_with_automl(
+        training_csv_path=".../training_spectra.csv",
+        feature_start_column="374.285004",
+        model_names=["RF", "SVR", "Ridge"],
+        n_trials=20,
+        timeout_sec=300,
+    )
+"""
+
+from __future__ import annotations
+
+import json
+import time
+from dataclasses import asdict, dataclass, field
+from pathlib import Path
+from typing import Any, Callable, Dict, List, Optional, Tuple
+
+import numpy as np
+import pandas as pd
+
+
+# ============================================================
+# 常量
+# ============================================================
+
+# AutoML 寻优阶段允许的最大样本数（避免 OOM）
+# 5000 样本对 RF/SVR/Ridge 的 Optuna 寻优足够给出稳定 CV
+DEFAULT_MAX_SAMPLES = 5000
+
+# 单次 Optuna trial 的默认超时（秒）
+DEFAULT_TIMEOUT = 300.0
+
+# 默认 trial 数
+DEFAULT_N_TRIALS = 20
+
+# AutoML 输出目录名后缀
+AUTOML_DIR_SUFFIX = "_AutoML"
+
+
+# ============================================================
+# 数据类
+# ============================================================
+
+@dataclass
+class AutoMLResult:
+    """单个目标列的 AutoML 训练结果"""
+    success: bool = False
+    model_path: Optional[str] = None
+    cv_score: float = -float("inf")
+    best_params: Optional[Dict[str, Any]] = None
+    target_column: str = ""
+    preprocessing: str = ""
+    model_name: str = ""
+    n_trials_done: int = 0
+    n_samples_used: int = 0
+    fallback_used: bool = False
+    elapsed_sec: float = 0.0
+    error: Optional[str] = None
+    metadata: Dict[str, Any] = field(default_factory=dict)
+
+
+# ============================================================
+# 智能子采样
+# ============================================================
+
+def smart_subsample(
+    X: np.ndarray,
+    y: np.ndarray,
+    max_samples: int = DEFAULT_MAX_SAMPLES,
+    random_state: int = 42,
+) -> Tuple[np.ndarray, np.ndarray, bool]:
+    """当 N > max_samples 时随机下采样；否则原样返回。
+
+    Returns:
+        (X_sub, y_sub, was_subsampled)
+    """
+    n = X.shape[0]
+    if n <= max_samples:
+        return X, y, False
+    rng = np.random.default_rng(random_state)
+    idx = rng.choice(n, size=max_samples, replace=False)
+    return X[idx], y[idx], True
+
+
+# ============================================================
+# 模型工厂
+# ============================================================
+
+def _build_model(model_name: str, random_state: int = 42):
+    """根据英文模型键名构造 sklearn-compatible 模型实例（factory）。"""
+    from sklearn.ensemble import (
+        AdaBoostRegressor, ExtraTreesRegressor, GradientBoostingRegressor,
+        RandomForestRegressor,
+    )
+    from sklearn.linear_model import (
+        ElasticNet, Lasso, LinearRegression, Ridge,
+    )
+    from sklearn.neighbors import KNeighborsRegressor
+    from sklearn.neural_network import MLPRegressor
+    from sklearn.svm import SVR
+    from sklearn.tree import DecisionTreeRegressor
+
+    factory = {
+        "RF": lambda **kw: RandomForestRegressor(random_state=random_state, n_jobs=1, **kw),
+        "ET": lambda **kw: ExtraTreesRegressor(random_state=random_state, n_jobs=1, **kw),
+        "GradientBoosting": lambda **kw: GradientBoostingRegressor(random_state=random_state, **kw),
+        "AdaBoost": lambda **kw: AdaBoostRegressor(random_state=random_state, **kw),
+        "Ridge": lambda **kw: Ridge(**kw),
+        "Lasso": lambda **kw: Lasso(max_iter=5000, **kw),
+        "ElasticNet": lambda **kw: ElasticNet(max_iter=5000, **kw),
+        "LinearRegression": lambda **kw: LinearRegression(**kw),
+        "SVR": lambda **kw: SVR(**kw),
+        "KNN": lambda **kw: KNeighborsRegressor(n_jobs=1, **kw),
+        "MLP": lambda **kw: MLPRegressor(max_iter=500, random_state=random_state, **kw),
+        "DecisionTree": lambda **kw: DecisionTreeRegressor(random_state=random_state, **kw),
+        "PLS": None,   # sklearn.cross_decomposition.PLSRegression 暂未集成
+    }
+    builder = factory.get(model_name)
+    if builder is None:
+        return None
+    return builder
+
+
+# ============================================================
+# Optuna 超参 search space
+# ============================================================
+
+def _get_search_space(model_name: str, trial) -> Dict[str, Any]:
+    """按模型名返回 Optuna 超参 search space。"""
+    sp: Dict[str, Any] = {}
+    if model_name == "RF":
+        sp["n_estimators"] = trial.suggest_int("n_estimators", 50, 300, step=50)
+        sp["max_depth"] = trial.suggest_int("max_depth", 3, 20)
+        sp["min_samples_split"] = trial.suggest_int("min_samples_split", 2, 10)
+        sp["min_samples_leaf"] = trial.suggest_int("min_samples_leaf", 1, 5)
+    elif model_name == "ET":
+        sp["n_estimators"] = trial.suggest_int("n_estimators", 50, 300, step=50)
+        sp["max_depth"] = trial.suggest_int("max_depth", 3, 20)
+    elif model_name == "GradientBoosting":
+        sp["n_estimators"] = trial.suggest_int("n_estimators", 50, 300, step=50)
+        sp["max_depth"] = trial.suggest_int("max_depth", 3, 8)
+        sp["learning_rate"] = trial.suggest_float("learning_rate", 0.01, 0.3, log=True)
+    elif model_name == "SVR":
+        sp["C"] = trial.suggest_float("C", 0.1, 100.0, log=True)
+        sp["epsilon"] = trial.suggest_float("epsilon", 0.001, 1.0, log=True)
+        sp["kernel"] = trial.suggest_categorical("kernel", ["rbf", "linear"])
+    elif model_name == "KNN":
+        sp["n_neighbors"] = trial.suggest_int("n_neighbors", 3, 20)
+        sp["weights"] = trial.suggest_categorical("weights", ["uniform", "distance"])
+    elif model_name in ("Ridge", "Lasso", "ElasticNet"):
+        sp["alpha"] = trial.suggest_float("alpha", 0.01, 100.0, log=True)
+        if model_name == "ElasticNet":
+            sp["l1_ratio"] = trial.suggest_float("l1_ratio", 0.0, 1.0)
+    elif model_name == "MLP":
+        sp["hidden_layer_sizes"] = trial.suggest_categorical(
+            "hidden_layer_sizes", [(50,), (100,), (50, 50), (100, 50)]
+        )
+        sp["alpha"] = trial.suggest_float("alpha", 1e-5, 1e-1, log=True)
+        sp["learning_rate_init"] = trial.suggest_float("learning_rate_init", 1e-4, 1e-2, log=True)
+    elif model_name == "DecisionTree":
+        sp["max_depth"] = trial.suggest_int("max_depth", 3, 20)
+        sp["min_samples_split"] = trial.suggest_int("min_samples_split", 2, 10)
+    elif model_name == "AdaBoost":
+        sp["n_estimators"] = trial.suggest_int("n_estimators", 30, 200, step=30)
+        sp["learning_rate"] = trial.suggest_float("learning_rate", 0.01, 1.0, log=True)
+    else:
+        sp["n_estimators"] = trial.suggest_int("n_estimators", 50, 200, step=50)
+    return sp
+
+
+def _make_objective(model_name: str, X: np.ndarray, y: np.ndarray,
+                    cv_folds: int, random_state: int):
+    """构造 Optuna objective（5 折 CV R²）。"""
+    from sklearn.model_selection import KFold, cross_val_score
+
+    def objective(trial):
+        params = _get_search_space(model_name, trial)
+        try:
+            builder = _build_model(model_name, random_state=random_state)
+            if builder is None:
+                return -1.0
+            model = builder(**params)
+            kf = KFold(n_splits=cv_folds, shuffle=True, random_state=random_state)
+            scores = cross_val_score(model, X, y, cv=kf, scoring="r2", n_jobs=1)
+            return float(np.mean(scores))
+        except Exception:
+            return -1.0
+
+    return objective
+
+
+def _refit_full(model_name: str, best_params: Dict[str, Any],
+                X: np.ndarray, y: np.ndarray, random_state: int):
+    """用 best params 在**全量数据**上 refit。"""
+    builder = _build_model(model_name, random_state=random_state)
+    if builder is None:
+        return None
+    model = builder(**best_params)
+    model.fit(X, y)
+    return model
+
+
+# ============================================================
+# 失败兜底（回退到老 GridSearchCV 路径）
+# ============================================================
+
+def _fallback_train(
+    training_csv_path: str,
+    feature_start_column,
+    preprocessing: str,
+    model_name: str,
+    split_method: str,
+    cv_folds: int,
+    output_dir: Path,
+    target_column: str,
+) -> AutoMLResult:
+    """AutoML 失败时调老 WaterQualityModelingBatch。
+
+    返回的 AutoMLResult.fallback_used=True。
+    """
+    try:
+        from src.core.modeling.modeling_batch import WaterQualityModelingBatch
+    except ImportError as e:
+        return AutoMLResult(
+            success=False, error=f"fallback 导入失败: {e!r}", fallback_used=True,
+            target_column=target_column, preprocessing=preprocessing, model_name=model_name,
+        )
+
+    try:
+        out_dir = output_dir / preprocessing
+        out_dir.mkdir(parents=True, exist_ok=True)
+        modeler = WaterQualityModelingBatch(str(out_dir))
+        modeler.train_models_batch(
+            csv_path=training_csv_path,
+            feature_start_column=feature_start_column,
+            preprocessing_methods=[preprocessing],
+            model_names=[model_name],
+            split_methods=[split_method],
+            cv_folds=cv_folds,
+        )
+        # 找产出
+        candidates = list(out_dir.rglob(f"{target_column}_{preprocessing}_{model_name}.joblib"))
+        model_path = str(candidates[0]) if candidates else None
+        return AutoMLResult(
+            success=model_path is not None,
+            model_path=model_path,
+            target_column=target_column, preprocessing=preprocessing, model_name=model_name,
+            fallback_used=True,
+            metadata={"source": "WaterQualityModelingBatch"},
+        )
+    except Exception as e:
+        return AutoMLResult(
+            success=False, error=f"fallback 失败: {e!r}", fallback_used=True,
+            target_column=target_column, preprocessing=preprocessing, model_name=model_name,
+        )
+
+
+# ============================================================
+# 主入口
+# ============================================================
+
+def train_with_automl(
+    training_csv_path: str,
+    feature_start_column,
+    preprocessing_methods: Optional[List[str]] = None,
+    model_names: Optional[List[str]] = None,
+    split_methods: Optional[List[str]] = None,
+    cv_folds: int = 5,
+    output_dir: Optional[str] = None,
+    n_trials: int = DEFAULT_N_TRIALS,
+    timeout_sec: float = DEFAULT_TIMEOUT,
+    max_samples: int = DEFAULT_MAX_SAMPLES,
+    random_state: int = 42,
+    callback: Optional[Callable[[str, str, str], None]] = None,
+) -> List[AutoMLResult]:
+    """用 Optuna + 子采样跑 AutoML。失败时自动回退到 GridSearchCV。
+
+    Args:
+        training_csv_path: 训练用 CSV（Step 5 产物 training_spectra.csv）
+        feature_start_column: 特征起始列名或索引（之前所有列视为目标 y）
+        preprocessing_methods: 候选预处理列表（**仅用第 1 个**，避免笛卡尔爆炸）
+        model_names: 候选模型列表（每个都会跑一遍 Optuna）
+        split_methods: 候选数据划分列表（AutoML 仅用第 1 个）
+        cv_folds: 交叉验证折数
+        output_dir: 输出目录（默认 <models_dir>_AutoML）
+        n_trials: 单模型 Optuna trial 数
+        timeout_sec: 单模型超时（秒），到时强制停止
+        max_samples: 寻优阶段允许的最大样本数
+        callback: 状态回调 callback(step_name, status, message)
+
+    Returns:
+        List[AutoMLResult]，每个目标列一份结果
+    """
+    def notify(status: str, msg: str = "") -> None:
+        if callback:
+            callback("步骤6_AutoML", status, msg)
+
+    # ---- 1) 参数默认值 ----
+    if preprocessing_methods is None:
+        preprocessing_methods = ["MMS"]
+    if model_names is None:
+        model_names = ["RF", "SVR", "Ridge"]
+    if split_methods is None:
+        split_methods = ["spxy"]
+
+    # 决策：仅用第一个预处理 + 第一个划分，避免笛卡尔爆炸
+    preproc = preprocessing_methods[0]
+    split_method = split_methods[0]
+
+    if output_dir is None:
+        output_dir = "./8_Supervised_Model_Training_AutoML"
+    out_dir = Path(output_dir)
+    out_dir.mkdir(parents=True, exist_ok=True)
+    preproc_dir = out_dir / preproc
+    preproc_dir.mkdir(parents=True, exist_ok=True)
+
+    # ---- 2) 加载数据 ----
+    notify("start", f"AutoML 训练开始 (n_trials={n_trials}, timeout={timeout_sec}s, max_samples={max_samples})")
+    if not Path(training_csv_path).exists():
+        return [AutoMLResult(success=False, error=f"训练 CSV 不存在: {training_csv_path}")]
+
+    df = pd.read_csv(training_csv_path)
+
+    # 提取目标列（feature_start_column 之前所有数值列）
+    if isinstance(feature_start_column, int):
+        y_cols = [c for c in df.columns[:feature_start_column]
+                  if pd.api.types.is_numeric_dtype(df[c])]
+    else:
+        try:
+            idx = list(df.columns).index(feature_start_column)
+            y_cols = [c for c in df.columns[:idx]
+                      if pd.api.types.is_numeric_dtype(df[c])]
+        except ValueError:
+            y_cols = []
+
+    if not y_cols:
+        notify("error", "AutoML: 未识别出目标列（feature_start_column 之前的所有数值列）")
+        return [AutoMLResult(success=False, error="未识别出目标列")]
+
+    feat_cols = [c for c in df.columns if c not in y_cols]
+    X_all = df[feat_cols].values.astype(np.float64)
+
+    # ---- 3) 预处理（仅第一项） ----
+    if preproc != "None":
+        try:
+            from src.preprocessing.spectral_Preprocessing import Preprocessing
+            processed = Preprocessing(preproc, df[feat_cols])
+            if isinstance(processed, pd.DataFrame):
+                X_all = processed.values.astype(np.float64)
+            else:
+                X_all = np.asarray(processed, dtype=np.float64)
+        except Exception as e:
+            notify("warning", f"预处理 {preproc} 失败: {e!r}，改用 None")
+            preproc = "None"
+
+    # ---- 4) 检查 Optuna 是否可用 ----
+    try:
+        import optuna
+        optuna.logging.set_verbosity(optuna.logging.WARNING)
+        optuna_available = True
+    except ImportError:
+        optuna_available = False
+        notify("warning", "optuna 未安装，全目标列回退到 GridSearchCV（pip install \"optuna>=3.6\"）")
+
+    # ---- 5) 逐 target 跑 ----
+    results: List[AutoMLResult] = []
+    total = len(y_cols)
+    per_model_timeout = max(10.0, timeout_sec / max(1, len(model_names)))
+
+    for ti, tgt in enumerate(y_cols, 1):
+        t0 = time.time()
+        yv = df[tgt].values.astype(np.float64)
+        mask = ~np.isnan(yv)
+        X_t = X_all[mask]
+        y_t = yv[mask]
+
+        if X_t.shape[0] < cv_folds * 2:
+            notify("warning", f"目标 {tgt}: 有效样本 {X_t.shape[0]} 不足，跳过")
+            results.append(AutoMLResult(
+                success=False, target_column=tgt, error=f"样本不足({X_t.shape[0]})",
+                preprocessing=preproc,
+            ))
+            continue
+
+        X_sub, y_sub, was_sub = smart_subsample(X_t, y_t, max_samples=max_samples, random_state=random_state)
+        if was_sub:
+            notify("info", f"目标 {tgt}: {X_t.shape[0]} 样本 → 子采样 {X_sub.shape[0]}（寻优用）")
+
+        best_overall = AutoMLResult(success=False, target_column=tgt, preprocessing=preproc)
+
+        if not optuna_available:
+            # 全目标列一次性 fallback
+            best_overall = _fallback_train(
+                training_csv_path, feature_start_column, preproc, model_names[0], split_method,
+                cv_folds, out_dir, tgt,
+            )
+        else:
+            for model_name in model_names:
+                try:
+                    builder = _build_model(model_name, random_state=random_state)
+                    if builder is None:
+                        notify("warning", f"模型 {model_name} 暂不支持 AutoML 寻优")
+                        continue
+
+                    study = optuna.create_study(
+                        direction="maximize",
+                        sampler=optuna.samplers.TPESampler(seed=random_state),
+                    )
+                    study.optimize(
+                        _make_objective(model_name, X_sub, y_sub, cv_folds, random_state),
+                        n_trials=n_trials,
+                        timeout=per_model_timeout,
+                        show_progress_bar=False,
+                    )
+
+                    if study.best_value is None or study.best_value <= -1.0:
+                        notify("warning", f"{tgt}/{model_name}: 全部 trial 失败（CV 全部 <= -1）")
+                        continue
+
+                    # refit on FULL
+                    final_model = _refit_full(model_name, study.best_params, X_t, y_t, random_state)
+                    if final_model is None:
+                        continue
+
+                    # 保存
+                    import joblib
+                    fname = f"{tgt}_{preproc}_{model_name}_AUTOML.joblib"
+                    fpath = preproc_dir / fname
+                    joblib.dump({
+                        "model": final_model,
+                        "target_column_name": tgt,
+                        "preprocess_method": preproc,
+                        "model_name": model_name,
+                        "metadata": {
+                            "automl": True,
+                            "best_params": study.best_params,
+                            "cv_score": float(study.best_value),
+                            "n_trials_done": len(study.trials),
+                            "n_samples_used_full": int(X_t.shape[0]),
+                            "n_samples_used_for_search": int(X_sub.shape[0]),
+                            "was_subsampled": was_sub,
+                            "split_method": split_method,
+                        },
+                    }, fpath)
+
+                    cand = AutoMLResult(
+                        success=True,
+                        model_path=str(fpath),
+                        cv_score=float(study.best_value),
+                        best_params=study.best_params,
+                        target_column=tgt,
+                        preprocessing=preproc,
+                        model_name=model_name,
+                        n_trials_done=len(study.trials),
+                        n_samples_used=int(X_sub.shape[0]),
+                        metadata={"refit_on_full": True, "n_samples_full": int(X_t.shape[0])},
+                    )
+                    if cand.cv_score > best_overall.cv_score:
+                        best_overall = cand
+                except Exception as e:
+                    notify("warning", f"目标 {tgt} / 模型 {model_name} 失败: {e!r}")
+                    continue
+
+            if not best_overall.success:
+                notify("warning", f"目标 {tgt} 全部 Optuna trial 失败，回退 GridSearchCV")
+                best_overall = _fallback_train(
+                    training_csv_path, feature_start_column, preproc, model_names[0], split_method,
+                    cv_folds, out_dir, tgt,
+                )
+
+        best_overall.elapsed_sec = time.time() - t0
+        results.append(best_overall)
+        notify("info", f"AutoML 目标 {tgt} 完成 ({ti}/{total}) cv={best_overall.cv_score:.4f}")
+
+    # ---- 6) 汇总 json ----
+    summary_path = out_dir / "automl_summary.json"
+    try:
+        with open(summary_path, "w", encoding="utf-8") as f:
+            json.dump([asdict(r) for r in results], f, ensure_ascii=False, indent=2, default=str)
+    except Exception as e:
+        notify("warning", f"写 automl_summary.json 失败: {e!r}")
+
+    success_n = sum(1 for r in results if r.success)
+    fallback_n = sum(1 for r in results if r.fallback_used)
+    notify("completed", f"AutoML 训练完成 {success_n}/{len(results)} 成功（{fallback_n} 走 fallback），汇总 {summary_path}")
+    return results
+
+
+# ============================================================
+# CLI 自测
+# ============================================================
+
+if __name__ == "__main__":
+    import argparse
+
+    p = argparse.ArgumentParser(description="AutoML 训练器 CLI 自测")
+    p.add_argument("--csv", required=True, help="训练用 CSV（feature_start_column 之前的列为目标 y）")
+    p.add_argument("--feature-start", default="0", help="特征起始列名或索引（默认 0）")
+    p.add_argument("--n-trials", type=int, default=DEFAULT_N_TRIALS)
+    p.add_argument("--timeout", type=float, default=DEFAULT_TIMEOUT)
+    p.add_argument("--max-samples", type=int, default=DEFAULT_MAX_SAMPLES)
+    p.add_argument("--out", default="./8_Supervised_Model_Training_AutoML")
+    args = p.parse_args()
+
+    # 智能推断 feature_start_column 类型
+    fsc: Any = args.feature_start
+    try:
+        fsc = int(fsc)
+    except ValueError:
+        pass
+
+    res = train_with_automl(
+        training_csv_path=args.csv,
+        feature_start_column=fsc,
+        n_trials=args.n_trials,
+        timeout_sec=args.timeout,
+        max_samples=args.max_samples,
+        output_dir=args.out,
+    )
+    print(f"\n训练完成 {len(res)} 个目标")
+    for r in res:
+        marker = "✓" if r.success else "✗"
+        fb = " [fallback]" if r.fallback_used else ""
+        print(f"  {marker} {r.target_column}: cv={r.cv_score:.4f} path={r.model_path}{fb}")

src/core/prediction/custom_regression_prediction.py

@@ -3,9 +3,9 @@
 """
 自定义回归预测模块
 
-该模块根据9_Custom_Regression_Modeling文件夹中的CSV信息，批量预测水质指数。
+该模块根据13_Custom_Regression文件夹中的CSV信息，批量预测水质指数。
 处理流程：
-1. 读取9_Custom_Regression_Modeling文件夹中的CSV文件
+1. 读取13_Custom_Regression文件夹中的CSV文件
 2. 根据r_squared选择最佳模型（指数公式+反演公式）
 3. 使用指数公式计算光谱指数值
 4. 使用反演公式计算水质参数值
@@ -38,12 +38,12 @@ class CustomRegressionPredictor:
     """
     自定义回归预测器
     
-    基于9_Custom_Regression_Modeling文件夹中的回归模型CSV文件，
+    基于13_Custom_Regression文件夹中的回归模型CSV文件，
     进行水质参数的批量预测。
     """
     
     def __init__(self, 
-                 regression_models_dir: str = "9_Custom_Regression_Modeling",
+                 regression_models_dir: str = "13_Custom_Regression",
                  formula_csv_path: Optional[str] = None,
                  output_dir: str = "prediction_results",
                  log_level: int = logging.INFO):
@@ -102,7 +102,7 @@ class CustomRegressionPredictor:
     
     def load_regression_models(self) -> Dict[str, pd.DataFrame]:
         """
-        加载9_Custom_Regression_Modeling文件夹中的所有CSV文件
+        加载13_Custom_Regression文件夹中的所有CSV文件
         
         支持的CSV格式：
         - 回归结果CSV包含列：y_variable, x_variable, equation, r_squared
@@ -621,7 +621,7 @@ def main():
     
     parser = argparse.ArgumentParser(description='自定义回归预测模块')
     parser.add_argument('--input_csv', required=True, help='输入的光谱采样CSV文件路径')
-    parser.add_argument('--models_dir', default='9_Custom_Regression_Modeling', 
+    parser.add_argument('--models_dir', default='13_Custom_Regression', 
                        help='回归模型CSV文件目录')
     parser.add_argument('--output_dir', default='prediction_results', 
                        help='预测结果输出目录')

src/core/prediction/inference_batch.py

@@ -13,6 +13,7 @@ import sys
 import os
 
 from src.preprocessing.spectral_Preprocessing import Preprocessing
+from src.core.utils.split_methods import spxy, ks
 
 # try:
 #     from modeling import WaterQualityModeling
@@ -26,19 +27,31 @@ from sklearn.model_selection import train_test_split
 class WaterQualityInference:
     """水质参数反演推理类"""
 
-    def __init__(self, artifacts_dir: str = "models/artifacts"):
+    def __init__(self, artifacts_dir: str = "models/artifacts",
+                 external_model=None, external_model_path=None):
         """
         初始化推理类
 
         Args:
             artifacts_dir: 模型保存目录
+            external_model: 外部预训练模型对象（来自 GUI 导入，跳过磁盘加载）
+            external_model_path: 外部模型文件路径（仅用于日志）
         """
         self.artifacts_dir = Path(artifacts_dir)
         if not self.artifacts_dir.exists():
             print(f"警告: 模型目录不存在: {artifacts_dir}，将在需要时创建")
-            
+
         self.best_model_info = None
-        self.loaded_model_data = None
+        self.external_model = external_model
+        self.external_model_path = external_model_path
+
+        # 规范化 loaded_model_data：始终为 dict，确保 ['model'] 访问不崩溃
+        if external_model is not None:
+            # 外部传入的是裸模型对象 → 包装为 dict，统一后续 .get('model') 访问
+            self.loaded_model_data = {'model': external_model, 'preprocess_method': 'None'}
+            print(f"  外部模型已规范化: type={type(external_model).__name__}")
+        else:
+            self.loaded_model_data = None
 
     def load_sampling_data(self, csv_path: str) -> Tuple[pd.DataFrame, pd.DataFrame, pd.DataFrame]:
         """
@@ -126,159 +139,12 @@ class WaterQualityInference:
         return X_train, X_test, y_train, y_test
 
     def spxy(self, data, label, test_size=0.2):
-        """
-        SPXY算法划分数据集（考虑X和Y空间的距离）
-        
-        Args:
-            data: shape (n_samples, n_features)
-            label: shape (n_samples, )
-            test_size: 测试集比例，默认: 0.2
-            
-        Returns:
-            X_train: (n_samples, n_features)
-            X_test: (n_samples, n_features)
-            y_train: (n_samples, )
-            y_test: (n_samples, )
-        """
-        # 确保 data 和 label 是 NumPy 数组
-        data = data.to_numpy() if isinstance(data, pd.DataFrame) else data
-        label = label.to_numpy() if isinstance(label, pd.Series) else label
-
-        # 备份原始数据和标签
-        x_backup = data
-        y_backup = label
-
-        M = data.shape[0]
-        N = round((1 - test_size) * M)
-        samples = np.arange(M)
-
-        # 归一化标签数据
-        label = (label - np.mean(label)) / np.std(label)
-        D = np.zeros((M, M))
-        Dy = np.zeros((M, M))
-
-        # 计算样本之间的距离
-        for i in range(M - 1):
-            xa = data[i, :]
-            ya = label[i]
-            for j in range((i + 1), M):
-                xb = data[j, :]
-                yb = label[j]
-                D[i, j] = np.linalg.norm(xa - xb)
-                Dy[i, j] = np.linalg.norm(ya - yb)
-
-        # 距离归一化
-        Dmax = np.max(D)
-        Dymax = np.max(Dy)
-        D = D / Dmax + Dy / Dymax
-
-        # 找到最远的两个点
-        maxD = D.max(axis=0)
-        index_row = D.argmax(axis=0)
-        index_column = maxD.argmax()
-
-        m = np.zeros(N, dtype=int)
-        m[0] = index_row[index_column]
-        m[1] = index_column
-
-        dminmax = np.zeros(N)
-        dminmax[1] = D[m[0], m[1]]
-
-        # 根据距离选择训练集
-        for i in range(2, N):
-            pool = np.delete(samples, m[:i])
-            dmin = np.zeros(M - i)
-            for j in range(M - i):
-                indexa = pool[j]
-                d = np.zeros(i)
-                for k in range(i):
-                    indexb = m[k]
-                    if indexa < indexb:
-                        d[k] = D[indexa, indexb]
-                    else:
-                        d[k] = D[indexb, indexa]
-                dmin[j] = np.min(d)
-            dminmax[i] = np.max(dmin)
-            index = np.argmax(dmin)
-            m[i] = pool[index]
-
-        m_complement = np.delete(samples, m)
-
-        # 划分训练集和测试集
-        X_train = data[m, :]
-        y_train = y_backup[m]
-        X_test = data[m_complement, :]
-        y_test = y_backup[m_complement]
-
-        return X_train, X_test, y_train, y_test
+        """SPXY算法划分数据集（委托至 src.core.utils.split_methods.spxy）"""
+        return spxy(data, label, test_size=test_size)
 
     def ks(self, data, label, test_size=0.2):
-        """
-        Kennard-Stone算法划分数据集
-        
-        Args:
-            data: shape (n_samples, n_features)
-            label: shape (n_sample, )
-            test_size: 测试集比例，默认: 0.2
-            
-        Returns:
-            X_train: (n_samples, n_features)
-            X_test: (n_samples, n_features)
-            y_train: (n_samples, )
-            y_test: (n_samples, )
-        """
-        # 确保 data 和 label 是 NumPy 数组
-        data = data.to_numpy() if isinstance(data, pd.DataFrame) else data
-        label = label.to_numpy() if isinstance(label, pd.Series) else label
-
-        M = data.shape[0]
-        N = round((1 - test_size) * M)
-        samples = np.arange(M)
-
-        D = np.zeros((M, M))
-
-        for i in range((M - 1)):
-            xa = data[i, :]
-            for j in range((i + 1), M):
-                xb = data[j, :]
-                D[i, j] = np.linalg.norm(xa - xb)
-
-        maxD = np.max(D, axis=0)
-        index_row = np.argmax(D, axis=0)
-        index_column = np.argmax(maxD)
-
-        m = np.zeros(N)
-        m[0] = np.array(index_row[index_column])
-        m[1] = np.array(index_column)
-        m = m.astype(int)
-        dminmax = np.zeros(N)
-        dminmax[1] = D[m[0], m[1]]
-
-        for i in range(2, N):
-            pool = np.delete(samples, m[:i])
-            dmin = np.zeros((M - i))
-            for j in range((M - i)):
-                indexa = pool[j]
-                d = np.zeros(i)
-                for k in range(i):
-                    indexb = m[k]
-                    if indexa < indexb:
-                        d[k] = D[indexa, indexb]
-                    else:
-                        d[k] = D[indexb, indexa]
-                dmin[j] = np.min(d)
-            dminmax[i] = np.max(dmin)
-            index = np.argmax(dmin)
-            m[i] = pool[index]
-
-        m_complement = np.delete(np.arange(data.shape[0]), m)
-
-        X_train = data[m, :]
-        y_train = label[m]
-        X_test = data[m_complement, :]
-        y_test = label[m_complement]
-
-        return X_train, X_test, y_train, y_test
+        """Kennard-Stone算法划分数据集（委托至 src.core.utils.split_methods.ks）"""
+        return ks(data, label, test_size=test_size)
 
     def split_data(self, X: np.ndarray, y: pd.Series, method: str = "random", 
                    test_size: float = 0.2, random_state: int = 42) -> Tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray]:
@@ -745,7 +611,10 @@ class WaterQualityInference:
             # 1. 加载模型
             print("\n步骤1: 加载模型")
             print("-" * 40)
-            if model_file_path:
+            if self.external_model is not None:
+                # 已在 __init__ 中规范化，无需重复赋值
+                print(f"  使用外部预训练模型: type={type(self.external_model).__name__}")
+            elif model_file_path:
                 self.load_specific_model(model_file_path)
             else:
                 self.load_best_model(metric=metric)
@@ -793,8 +662,8 @@ class WaterQualityInference:
 
         info = {
             "status": "model_loaded",
-            "preprocess_method": self.loaded_model_data['preprocess_method'],
-            "model_name": self.loaded_model_data['model_name'],
+            "preprocess_method": self.loaded_model_data.get('preprocess_method', 'Unknown'),
+            "model_name": self.loaded_model_data.get('model_name', type(self.external_model).__name__ if self.external_model else 'Unknown'),
             "model_type": str(type(self.loaded_model_data['model'])),
             "metadata": self.loaded_model_data.get('metadata', {})
         }
@@ -863,10 +732,13 @@ class WaterQualityInference:
         print(f"\n批量推理完成，共处理 {len(csv_files)} 个文件")
         return results
 
-    def batch_inference_multi_models(self, models_root_dir: str, sampling_csv_path: str, 
-                                   output_dir: str, metric: str = 'test_r2', 
+    def batch_inference_multi_models(self, models_root_dir: str, sampling_csv_path: str,
+                                   output_dir: str, metric: str = 'test_r2',
                                    prediction_column: str = 'prediction',
-                                   output_format: str = 'csv'):
+                                   output_format: str = 'csv',
+                                   external_model=None,
+                                   external_model_path=None,
+                                   external_models_dict=None):
         """
         使用多个子文件夹中的模型进行批量推理
         
@@ -881,28 +753,62 @@ class WaterQualityInference:
         models_root = Path(models_root_dir)
         output_path = Path(output_dir)
         output_path.mkdir(parents=True, exist_ok=True)
-        
-        # 查找所有子文件夹
-        subdirs = [d for d in models_root.iterdir() if d.is_dir()]
-        
-        if not subdirs:
-            print(f"在目录 {models_root_dir} 中未找到子文件夹")
-            return
-            
-        print(f"找到 {len(subdirs)} 个模型子文件夹进行批量推理")
-        print(f"输出格式: {output_format.upper()}")
-        
+
         all_results = {}
-        
-        for subdir in subdirs:
+
+        # 优先级 1：_external_models_dict 非空 → 直接用字典的 keys 作为 targets，不扫描磁盘
+        print(f"[BatchInference] 终于收到字典啦！包含模型: {list(external_models_dict.keys()) if external_models_dict else 'None'}")
+        if external_models_dict is not None and len(external_models_dict) > 0:
+            targets = list(external_models_dict.keys())
+            print(f"\n使用外部导入模型字典（{len(targets)} 个模型）")
+            print(f"检测到外部导入模型，将预测以下参数: {targets}")
+        elif external_model is not None:
+            print(f"\n使用外部预训练模型: {external_model_path or 'unknown'}")
+            subdirs = [d for d in models_root.iterdir() if d.is_dir()]
+            if not subdirs:
+                print(f"在目录 {models_root_dir} 中未找到子文件夹")
+                return {}
+            print(f"找到 {len(subdirs)} 个模型子文件夹进行批量推理")
+            targets = [d.name for d in subdirs]
+        else:
+            subdirs = [d for d in models_root.iterdir() if d.is_dir()]
+            if not subdirs:
+                print(f"在目录 {models_root_dir} 中未找到子文件夹")
+                return {}
+            print(f"找到 {len(subdirs)} 个模型子文件夹进行批量推理")
+            targets = [d.name for d in subdirs]
+
+        print(f"输出格式: {output_format.upper()}")
+
+        for subdir_name in targets:
             try:
-                subdir_name = subdir.name
                 print(f"\n{'='*60}")
-                print(f"处理模型文件夹: {subdir_name}")
+                print(f"处理模型: {subdir_name}")
                 print(f"{'='*60}")
-                
-                # 创建新的推理实例，使用当前子文件夹作为artifacts_dir
-                model_inferencer = WaterQualityInference(str(subdir))
+
+                # 优先级：字典中该 target 的模型 > 共享单模型 > 磁盘加载
+                effective_model = None
+                if external_models_dict and subdir_name in external_models_dict:
+                    effective_model = external_models_dict[subdir_name]
+                    print(f"  → 使用字典中模型: {type(effective_model).__name__}")
+                elif external_model is not None:
+                    effective_model = external_model
+                    print(f"  → 使用共享外部模型: {type(effective_model).__name__}")
+
+                # artifacts_dir：字典模式优先用 placeholder "./"，否则用真实子目录
+                artifacts_dir = (
+                    str(models_root / subdir_name)
+                    if (models_root / subdir_name).is_dir()
+                    else str(models_root)
+                )
+                if effective_model is not None:
+                    model_inferencer = WaterQualityInference(
+                        artifacts_dir,
+                        external_model=effective_model,
+                        external_model_path=external_model_path or "",
+                    )
+                else:
+                    model_inferencer = WaterQualityInference(artifacts_dir)
                 
                 # 根据输出格式设置文件扩展名
                 file_ext = f".{output_format}"
@@ -931,10 +837,10 @@ class WaterQualityInference:
                     }
                 }
                 
-                print(f"子文件夹 {subdir_name} 处理完成")
+                print(f"模型 {subdir_name} 处理完成")
                 
             except Exception as e:
-                print(f"处理子文件夹 {subdir_name} 失败: {e}")
+                print(f"处理模型 {subdir_name} 失败: {e}")
                 all_results[subdir_name] = {
                     'status': 'error',
                     'error': str(e)

src/core/prediction/sctter_batch.py

@@ -24,6 +24,7 @@ from sklearn.linear_model import LinearRegression, Ridge, Lasso, ElasticNet
 from sklearn.model_selection import GridSearchCV, cross_val_score, KFold, train_test_split
 from sklearn.metrics import mean_squared_error, mean_absolute_error, r2_score
 from sklearn.cross_decomposition import PLSRegression
+from src.core.utils.split_methods import spxy, ks
 
 # 第三方模型导入
 # try:
@@ -256,133 +257,12 @@ class WaterQualityScatterBatch:
         return X_train, X_test, y_train, y_test
 
     def spxy(self, data, label, test_size=0.2):
-        """SPXY算法划分数据集"""
-        # 确保 data 和 label 是 NumPy 数组
-        data = data.to_numpy() if isinstance(data, pd.DataFrame) else data
-        label = label.to_numpy() if isinstance(label, pd.Series) else label
-
-        # 备份原始数据和标签
-        x_backup = data
-        y_backup = label
-
-        M = data.shape[0]
-        N = round((1 - test_size) * M)
-        samples = np.arange(M)
-
-        # 归一化标签数据
-        label = (label - np.mean(label)) / np.std(label)
-        D = np.zeros((M, M))
-        Dy = np.zeros((M, M))
-
-        # 计算样本之间的距离
-        for i in range(M - 1):
-            xa = data[i, :]
-            ya = label[i]
-            for j in range((i + 1), M):
-                xb = data[j, :]
-                yb = label[j]
-                D[i, j] = np.linalg.norm(xa - xb)
-                Dy[i, j] = np.linalg.norm(ya - yb)
-
-        # 距离归一化
-        Dmax = np.max(D)
-        Dymax = np.max(Dy)
-        D = D / Dmax + Dy / Dymax
-
-        # 找到最远的两个点
-        maxD = D.max(axis=0)
-        index_row = D.argmax(axis=0)
-        index_column = maxD.argmax()
-
-        m = np.zeros(N, dtype=int)
-        m[0] = index_row[index_column]
-        m[1] = index_column
-
-        dminmax = np.zeros(N)
-        dminmax[1] = D[m[0], m[1]]
-
-        # 根据距离选择训练集
-        for i in range(2, N):
-            pool = np.delete(samples, m[:i])
-            dmin = np.zeros(M - i)
-            for j in range(M - i):
-                indexa = pool[j]
-                d = np.zeros(i)
-                for k in range(i):
-                    indexb = m[k]
-                    if indexa < indexb:
-                        d[k] = D[indexa, indexb]
-                    else:
-                        d[k] = D[indexb, indexa]
-                dmin[j] = np.min(d)
-            dminmax[i] = np.max(dmin)
-            index = np.argmax(dmin)
-            m[i] = pool[index]
-
-        m_complement = np.delete(samples, m)
-
-        # 划分训练集和测试集
-        X_train = data[m, :]
-        y_train = y_backup[m]
-        X_test = data[m_complement, :]
-        y_test = y_backup[m_complement]
-
-        return X_train, X_test, y_train, y_test
+        """SPXY算法划分数据集（委托至 src.core.utils.split_methods.spxy）"""
+        return spxy(data, label, test_size=test_size)
 
     def ks(self, data, label, test_size=0.2):
-        """Kennard-Stone算法划分数据集"""
-        # 确保 data 和 label 是 NumPy 数组
-        data = data.to_numpy() if isinstance(data, pd.DataFrame) else data
-        label = label.to_numpy() if isinstance(label, pd.Series) else label
-
-        M = data.shape[0]
-        N = round((1 - test_size) * M)
-        samples = np.arange(M)
-
-        D = np.zeros((M, M))
-
-        for i in range((M - 1)):
-            xa = data[i, :]
-            for j in range((i + 1), M):
-                xb = data[j, :]
-                D[i, j] = np.linalg.norm(xa - xb)
-
-        maxD = np.max(D, axis=0)
-        index_row = np.argmax(D, axis=0)
-        index_column = np.argmax(maxD)
-
-        m = np.zeros(N)
-        m[0] = np.array(index_row[index_column])
-        m[1] = np.array(index_column)
-        m = m.astype(int)
-        dminmax = np.zeros(N)
-        dminmax[1] = D[m[0], m[1]]
-
-        for i in range(2, N):
-            pool = np.delete(samples, m[:i])
-            dmin = np.zeros((M - i))
-            for j in range((M - i)):
-                indexa = pool[j]
-                d = np.zeros(i)
-                for k in range(i):
-                    indexb = m[k]
-                    if indexa < indexb:
-                        d[k] = D[indexa, indexb]
-                    else:
-                        d[k] = D[indexb, indexa]
-                dmin[j] = np.min(d)
-            dminmax[i] = np.max(dmin)
-            index = np.argmax(dmin)
-            m[i] = pool[index]
-
-        m_complement = np.delete(np.arange(data.shape[0]), m)
-
-        X_train = data[m, :]
-        y_train = label[m]
-        X_test = data[m_complement, :]
-        y_test = label[m_complement]
-
-        return X_train, X_test, y_train, y_test
+        """Kennard-Stone算法划分数据集（委托至 src.core.utils.split_methods.ks）"""
+        return ks(data, label, test_size=test_size)
 
     def split_data(self, X: np.ndarray, y: pd.Series, method: str = "random", 
                    test_size: float = 0.2, random_state: int = 42) -> Tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray]:

src/core/steps/data_preparation_step.py

@@ -2,7 +2,7 @@
 """
 数据准备步骤
 
-包含 step4_process_csv, step5_extract_training_spectra, step5_5_calculate_water_quality_indices
+包含 step5_process_csv, step6_extract_spectra, step5_5_calculate_water_quality_indices
 """
 
 import time
@@ -21,7 +21,7 @@ class DataPreparationStep:
     @staticmethod
     def process_csv(
         csv_path: str,
-        output_dir: Union[str, Path] = "./4_processed_data",
+        output_dir: Union[str, Path] = "./5_Data_Cleaning",
         callback: Optional[Callable] = None,
     ) -> str:
         """处理CSV文件（筛选剔除异常值）"""
@@ -61,7 +61,7 @@ class DataPreparationStep:
         boundary_path: Optional[str] = None,
         glint_mask_path: Optional[str] = None,
         water_mask_path: Optional[str] = None,
-        output_dir: Union[str, Path] = "./5_training_spectra",
+        output_dir: Union[str, Path] = "./6_Spectral_Feature_Extraction",
         callback: Optional[Callable] = None,
     ) -> str:
         """根据采样点坐标在去耀斑影像中提取平均光谱"""
@@ -126,12 +126,12 @@ class DataPreparationStep:
 
     @staticmethod
     def calculate_water_quality_indices(
-        training_spectra_path: Optional[str] = None,
+        training_csv_path: Optional[str] = None,
         formula_csv_file: Optional[str] = None,
         formula_names: Optional[List[str]] = None,
         output_file: Optional[str] = None,
         enabled: bool = True,
-        output_dir: Union[str, Path] = "./6_water_quality_indices",
+        output_dir: Union[str, Path] = "./7_Water_Quality_Indices",
         callback: Optional[Callable] = None,
     ) -> Optional[str]:
         """根据训练光谱计算水质光谱指数（使用 band_math 方法）"""
@@ -153,15 +153,15 @@ class DataPreparationStep:
             notify("skipped", "跳过水质指数计算")
             return None
 
-        if training_spectra_path is None:
-            raise ValueError("必须提供 training_spectra_path 参数")
+        if training_csv_path is None:
+            raise ValueError("必须提供 training_csv_path 参数")
         if formula_csv_file is None:
             raise ValueError("必须提供 formula_csv_file 参数")
 
         if output_file:
             output_path = str(Path(output_file))
         else:
-            output_path = str(output_dir / "water_quality_indices.csv")
+            output_path = str(output_dir / "training_spectra_indices.csv")
 
         if Path(output_path).exists():
             print(f"检测到已存在的水质指数文件，直接使用: {output_path}")
@@ -170,7 +170,7 @@ class DataPreparationStep:
 
         from src.utils.band_math import BandMathCalculator
 
-        calculator = BandMathCalculator(training_spectra_path)
+        calculator = BandMathCalculator(training_csv_path)
         result_df = calculator.process_formulas_from_csv(
             formula_csv_file=formula_csv_file,
             formula_names=formula_names,

src/core/steps/glint_detection_step.py

@@ -28,7 +28,7 @@ class GlintDetectionStep:
         max_area: Optional[int] = None,
         buffer_size: Optional[int] = None,
         water_mask_path: Optional[str] = None,
-        glint_dir: Union[str, Path] = "./2_glint",
+        glint_dir: Union[str, Path] = "./2_Glint_Detection",
         callback: Optional[callable] = None,
     ) -> str:
         """

src/core/steps/modeling_step.py

@@ -135,7 +135,7 @@ class ModelingStep:
         split_methods: Optional[List[str]] = None,
         cv_folds: int = 5,
         training_csv_path: Optional[str] = None,
-        output_dir: Union[str, Path] = "./7_Supervised_Model_Training",
+        output_dir: Union[str, Path] = "./8_Supervised_Model_Training",
         callback: Optional[Callable] = None,
         _report_generator=None,
     ) -> str:
@@ -251,7 +251,7 @@ class ModelingStep:
         if output_dir is not None:
             non_empirical_dir = Path(output_dir)
         else:
-            non_empirical_dir = Path.cwd() / "8_Regression_Modeling"
+            non_empirical_dir = Path.cwd() / "8_Non_Empirical_Regression"
         non_empirical_dir.mkdir(parents=True, exist_ok=True)
 
         if preprocessing_methods is None:
@@ -362,7 +362,7 @@ class ModelingStep:
             raise ValueError(f"因变量列不存在: {missing_y}")
 
         if output_dir is None:
-            custom_regression_dir = Path(work_dir) / "9_Custom_Regression_Modeling"
+            custom_regression_dir = Path(work_dir) / "13_Custom_Regression"
         else:
             custom_regression_dir = Path(work_dir) / output_dir
         custom_regression_dir.mkdir(parents=True, exist_ok=True)
@@ -430,7 +430,7 @@ def _apply_preprocessing_internal(
 
     save_path = None
     if preprocess_method == "SS":
-        models_dir = output_dir.parent.parent / "7_Supervised_Model_Training"
+        models_dir = output_dir.parent.parent / "8_Supervised_Model_Training"
         models_dir.mkdir(parents=True, exist_ok=True)
         save_path = str(models_dir / "scaler_params.pkl")
         print(f"SS预处理: scaler模型将保存到 {save_path}")

src/core/steps/prediction_step.py

@@ -24,8 +24,9 @@ class PredictionStep:
         chunk_size: int = 1000,
         water_mask_path: Optional[str] = None,
         glint_mask_path: Optional[str] = None,
-        output_dir: Union[str, Path] = "./10_sampling",
+        output_dir: Union[str, Path] = "./4_sampling",
         callback: Optional[Callable] = None,
+        use_adaptive_sampling: bool = True,
     ) -> str:
         """生成水域掩膜内且耀斑掩膜外的采样点，统计平均光谱"""
         from pathlib import Path
@@ -83,10 +84,14 @@ class PredictionStep:
         if glint_mask_to_use is None:
             print("未检测到耀斑掩膜，将在采样点生成时不做耀斑区域剔除。")
 
-        # 传递极度安全的 deglint_img_str 进底层
+        # 传递极度安全的 deglint_img_str 进底层（关键字传参，避免 positional 参数顺序陷阱）
         get_spectral_sampling_points_chunked(
             deglint_img_str, water_mask_path, glint_mask_to_use,
-            output_path, interval, sample_radius, chunk_size
+            output_path,
+            interval=interval,
+            sample_radius=sample_radius,
+            chunk_size=chunk_size,
+            use_adaptive_sampling=use_adaptive_sampling,
         )
 
         notify("completed", f"采样点光谱数据已保存: {output_path}")
@@ -100,9 +105,13 @@ class PredictionStep:
         models_dir: Optional[str] = None,
         metric: str = "test_r2",
         prediction_column: str = "prediction",
-        output_dir: Union[str, Path] = "./11_12_13_predictions/Machine_Learning_Prediction",
+        output_dir: Union[str, Path] = "./9_ML_Prediction",
         callback: Optional[Callable] = None,
         _report_generator=None,
+        _external_model=None,
+        _external_model_path=None,
+        _external_models_dict=None,
+        _external_model_dir=None,
     ) -> Dict[str, str]:
         """将训练好的最佳机器学习模型应用到采样点光谱上，预测水质参数"""
         from src.core.prediction.inference_batch import WaterQualityInference
@@ -114,6 +123,8 @@ class PredictionStep:
         print("\n" + "=" * 80)
         print("步骤8: 预测水质参数")
         print("=" * 80)
+        print(f"[PredictionStep] 准备执行预测，字典状态: {'Yes' if _external_models_dict else 'No'}"
+              f", 单模型状态: {'Yes' if _external_model else 'No'}")
 
         step_start_time = time.time()
 
@@ -149,19 +160,60 @@ class PredictionStep:
                 else:
                     print(f"检测到部分预测结果文件，缺少: {missing_targets}，将继续生成...")
 
-        inferencer = WaterQualityInference(models_dir)
-        all_results = inferencer.batch_inference_multi_models(
-            models_root_dir=models_dir,
-            sampling_csv_path=sampling_csv_path,
-            output_dir=str(ml_prediction_dir),
-            metric=metric,
-            prediction_column=prediction_column,
-            output_format="csv",
-        )
+        all_results = {}
 
-        for target_name, result in all_results.items():
-            if result.get("status") == "success":
-                prediction_files[target_name] = result["output_file"]
+        if _external_models_dict:
+            # 外部模型字典优先：直接用字典的 keys 作为 targets 列表，
+            # 手动为每个模型创建 inference 实例并调用 inference_pipeline。
+            print(f"\n使用外部导入模型字典（{len(_external_models_dict)} 个模型）...")
+            for target_name, model_obj in _external_models_dict.items():
+                try:
+                    output_file = ml_prediction_dir / f"{target_name}.csv"
+                    model_inferencer = WaterQualityInference(
+                        models_dir or "./",
+                        external_model=model_obj,
+                        external_model_path=_external_model_dir or "",
+                    )
+                    predictions, result_df = model_inferencer.inference_pipeline(
+                        sampling_csv_path=sampling_csv_path,
+                        output_csv_path=str(output_file),
+                        metric=metric,
+                        prediction_column=prediction_column,
+                    )
+                    prediction_files[target_name] = str(output_file)
+                    all_results[target_name] = {
+                        "status": "success",
+                        "output_file": str(output_file),
+                        "sample_count": len(predictions),
+                    }
+                    print(f"  ✓ {target_name}: {len(predictions)} 个预测值")
+                except Exception as e:
+                    print(f"  ✗ {target_name}: 失败 — {type(e).__name__}: {e}")
+                    prediction_files[target_name] = None
+                    all_results[target_name] = {"status": "error", "error": str(e)}
+        else:
+            # 字典为空或不存在：回退到扫描 models_dir 子目录的传统逻辑
+            inferencer = WaterQualityInference(
+                models_dir,
+                external_model=_external_model,
+                external_model_path=_external_model_path,
+            )
+            all_results = inferencer.batch_inference_multi_models(
+                models_root_dir=models_dir,
+                sampling_csv_path=sampling_csv_path,
+                output_dir=str(ml_prediction_dir),
+                metric=metric,
+                prediction_column=prediction_column,
+                output_format="csv",
+                external_model=_external_model,
+                external_model_path=_external_model_path,
+                external_models_dict=_external_models_dict,
+            )
+            # batch_inference_multi_models 已确保返回字典，永不返回 None
+            if all_results:
+                for target_name, result in all_results.items():
+                    if result.get("status") == "success":
+                        prediction_files[target_name] = result["output_file"]
 
         print(f"预测完成，结果保存在: {ml_prediction_dir}")
 
@@ -207,7 +259,7 @@ class PredictionStep:
         if non_empirical_models_dir is not None:
             final_models_dir = non_empirical_models_dir
         else:
-            default_models_dir = str(Path(work_dir) / "8_Regression_Modeling")
+            default_models_dir = str(Path(work_dir) / "8_Non_Empirical_Regression")
             if Path(default_models_dir).exists():
                 final_models_dir = default_models_dir
             else:
@@ -311,14 +363,14 @@ class PredictionStep:
         if custom_regression_dir is not None:
             final_regression_dir = custom_regression_dir
         else:
-            final_regression_dir = str(Path(work_dir) / "9_Custom_Regression_Modeling")
+            final_regression_dir = str(Path(work_dir) / "13_Custom_Regression")
             if not Path(final_regression_dir).exists():
                 raise ValueError(
                     "请先执行步骤6.75: 自定义回归分析，或提供 custom_regression_dir 参数"
                 )
 
         if output_dir is None:
-            custom_regression_prediction_dir = Path(work_dir) / "11_12_13_predictions" / "Custom_Regression_Prediction"
+            custom_regression_prediction_dir = Path(work_dir) / "13_Custom_Regression" / "Custom_Regression_Prediction"
             custom_regression_prediction_dir.mkdir(parents=True, exist_ok=True)
             prediction_output_dir = str(custom_regression_prediction_dir)
         else:

src/core/utils/split_methods.py

@@ -0,0 +1,158 @@
+# -*- coding: utf-8 -*-
+"""
+数据集划分算法 —— SPXY / Kennard-Stone
+
+从 modeling_batch.py / inference_batch.py / sctter_batch.py 中抽离，
+消除三处完全相同的重复实现。
+"""
+
+import numpy as np
+import pandas as pd
+
+
+def spxy(data, label, test_size=0.2):
+    """
+    SPXY算法划分数据集（考虑X和Y空间的距离）
+
+    Args:
+        data: shape (n_samples, n_features) —— np.ndarray 或 pd.DataFrame
+        label: shape (n_samples, ) —— np.ndarray 或 pd.Series
+        test_size: 测试集比例，默认: 0.2
+
+    Returns:
+        X_train: (n_samples, n_features)
+        X_test: (n_samples, n_features)
+        y_train: (n_samples, )
+        y_test: (n_samples, )
+    """
+    data = data.to_numpy() if isinstance(data, pd.DataFrame) else data
+    label = label.to_numpy() if isinstance(label, pd.Series) else label
+
+    x_backup = data
+    y_backup = label
+
+    M = data.shape[0]
+    N = round((1 - test_size) * M)
+    samples = np.arange(M)
+
+    label = (label - np.mean(label)) / np.std(label)
+    D = np.zeros((M, M))
+    Dy = np.zeros((M, M))
+
+    for i in range(M - 1):
+        xa = data[i, :]
+        ya = label[i]
+        for j in range((i + 1), M):
+            xb = data[j, :]
+            yb = label[j]
+            D[i, j] = np.linalg.norm(xa - xb)
+            Dy[i, j] = np.linalg.norm(ya - yb)
+
+    Dmax = np.max(D)
+    Dymax = np.max(Dy)
+    D = D / Dmax + Dy / Dymax
+
+    maxD = D.max(axis=0)
+    index_row = D.argmax(axis=0)
+    index_column = maxD.argmax()
+
+    m = np.zeros(N, dtype=int)
+    m[0] = index_row[index_column]
+    m[1] = index_column
+
+    dminmax = np.zeros(N)
+    dminmax[1] = D[m[0], m[1]]
+
+    for i in range(2, N):
+        pool = np.delete(samples, m[:i])
+        dmin = np.zeros(M - i)
+        for j in range(M - i):
+            indexa = pool[j]
+            d = np.zeros(i)
+            for k in range(i):
+                indexb = m[k]
+                if indexa < indexb:
+                    d[k] = D[indexa, indexb]
+                else:
+                    d[k] = D[indexb, indexa]
+            dmin[j] = np.min(d)
+        dminmax[i] = np.max(dmin)
+        index = np.argmax(dmin)
+        m[i] = pool[index]
+
+    m_complement = np.delete(samples, m)
+
+    X_train = data[m, :]
+    y_train = y_backup[m]
+    X_test = data[m_complement, :]
+    y_test = y_backup[m_complement]
+
+    return X_train, X_test, y_train, y_test
+
+
+def ks(data, label, test_size=0.2):
+    """
+    Kennard-Stone算法划分数据集
+
+    Args:
+        data: shape (n_samples, n_features) —— np.ndarray 或 pd.DataFrame
+        label: shape (n_samples, ) —— np.ndarray 或 pd.Series
+        test_size: 测试集比例，默认: 0.2
+
+    Returns:
+        X_train: (n_samples, n_features)
+        X_test: (n_samples, n_features)
+        y_train: (n_samples, )
+        y_test: (n_samples, )
+    """
+    data = data.to_numpy() if isinstance(data, pd.DataFrame) else data
+    label = label.to_numpy() if isinstance(label, pd.Series) else label
+
+    M = data.shape[0]
+    N = round((1 - test_size) * M)
+    samples = np.arange(M)
+
+    D = np.zeros((M, M))
+
+    for i in range((M - 1)):
+        xa = data[i, :]
+        for j in range((i + 1), M):
+            xb = data[j, :]
+            D[i, j] = np.linalg.norm(xa - xb)
+
+    maxD = np.max(D, axis=0)
+    index_row = np.argmax(D, axis=0)
+    index_column = np.argmax(maxD)
+
+    m = np.zeros(N)
+    m[0] = np.array(index_row[index_column])
+    m[1] = np.array(index_column)
+    m = m.astype(int)
+    dminmax = np.zeros(N)
+    dminmax[1] = D[m[0], m[1]]
+
+    for i in range(2, N):
+        pool = np.delete(samples, m[:i])
+        dmin = np.zeros((M - i))
+        for j in range((M - i)):
+            indexa = pool[j]
+            d = np.zeros(i)
+            for k in range(i):
+                indexb = m[k]
+                if indexa < indexb:
+                    d[k] = D[indexa, indexb]
+                else:
+                    d[k] = D[indexb, indexa]
+            dmin[j] = np.min(d)
+        dminmax[i] = np.max(dmin)
+        index = np.argmax(dmin)
+        m[i] = pool[index]
+
+    m_complement = np.delete(np.arange(data.shape[0]), m)
+
+    X_train = data[m, :]
+    y_train = label[m]
+    X_test = data[m_complement, :]
+    y_test = label[m_complement]
+
+    return X_train, X_test, y_train, y_test

src/core/visualization/preview.py

@@ -16,12 +16,12 @@ def generate_glint_deglint_previews(
     output_dir: Optional[str] = None
 ) -> Dict[str, str]:
     """
-    生成2_glint和3_deglint文件夹中影像文件的PNG预览图
+    生成2_Glint_Detection和3_deglint文件夹中影像文件的PNG预览图
 
     Args:
         work_dir: 工作目录
         output_subdir: 输出子目录名称
-        generate_glint: 是否处理2_glint文件夹
+        generate_glint: 是否处理2_Glint_Detection文件夹
         generate_deglint: 是否处理3_deglint文件夹
         output_dir: 输出目录（None则使用默认）
 

src/core/workspace_manager.py

@@ -0,0 +1,283 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+工作空间管理器
+
+负责工作目录文件扫描、步骤输出路径发现、配置裁剪等业务逻辑，
+与 GUI 组件解耦，不直接引用任何 UI 类。
+"""
+
+import copy
+from pathlib import Path
+
+from src.gui.core.event_bus import global_event_bus
+
+
+class WorkspaceManager:
+    """管理步骤默认输出路径、文件扫描与配置裁剪"""
+
+    # 白名单：科学数据格式后缀
+    SCIENTIFIC_EXTENSIONS = {'.dat', '.tif', '.tiff', '.shp'}
+    # 临时文件关键词黑名单
+    TMP_KEYWORDS = ('__tmp', '_tmp')
+    # 掩膜类型集合
+    MASK_TYPES = {'water_mask', 'glint_mask', 'boundary_mask'}
+
+    def __init__(self):
+        self.step_default_outputs = {
+            'step1': {'water_mask': [
+                "1_water_mask/water_mask_out.dat",
+                "1_water_mask/water_mask_from_ndwi.dat",
+                "1_water_mask/water_mask_from_shp.dat",
+            ]},
+            'step2': {'glint_mask': "2_Glint_Detection/severe_glint_area.dat"},
+            'step3': {'deglint_image': [
+                "3_deglint/deglint_image.bsq",
+                "3_deglint/deglint_goodman.bsq",
+            ]},
+            'step4_sampling': {'sampling_points': "4_sampling/sampling_spectra.csv"},
+            'step5_clean': {'processed_data': "5_Data_Cleaning/processed_data.csv"},
+            'step6_feature': {'training_spectra': "6_Spectral_Feature_Extraction/training_spectra.csv"},
+            'step7_index': {'training_spectra_indices': "7_Water_Quality_Indices/training_spectra_indices.csv"},
+            'step8_ml_train': {'Supervised_Model_Training': "8_Supervised_Model_Training/"},
+            'step9_ml_predict': {'9_ML_Prediction': "9_ML_Prediction/"},
+            'step10_watercolor': {'WaterIndex_Images': "10_WaterIndex_Images/"},
+            'step11_map': {'14_visualization': "14_visualization/"},
+        }
+        self.step_outputs = {}
+
+    def _publish_outputs(self, step_id: str, outputs: dict):
+        """将发现的产出发布到 EventBus。
+
+        Args:
+            step_id: 面板 step_id（如 'step1', 'step5_clean'）
+            outputs: {output_type: path_str}
+        """
+        for output_type, path in outputs.items():
+            if path:
+                global_event_bus.publish('OutputUpdated', {
+                    'step_id': step_id,
+                    'output_type': output_type,
+                    'path': path,
+                })
+
+    @staticmethod
+    def _is_scientific_mask(path_str):
+        """白名单判断：只有 .dat .tif .tiff .shp 才算科学数据格式"""
+        p = Path(path_str)
+        name_lower = str(path_str).lower()
+        if any(kw in name_lower for kw in WorkspaceManager.TMP_KEYWORDS):
+            return False
+        return p.suffix.lower() in WorkspaceManager.SCIENTIFIC_EXTENSIONS
+
+    def find_step_output(self, work_path, step_id, output_type, ref_img_path=None):
+        """查找指定步骤的输出文件
+
+        Args:
+            work_path: 工作目录 Path 对象
+            step_id: 步骤 ID
+            output_type: 输出类型（如 'water_mask', 'deglint_image' 等）
+            ref_img_path: 参考影像路径（仅 output_type='reference_img' 时需要）
+
+        Returns:
+            找到的文件路径字符串，或 None
+        """
+        if step_id not in self.step_default_outputs:
+            return None
+
+        raw = self.step_default_outputs[step_id]
+
+        rel_path = None
+        if isinstance(raw, str):
+            rel_path = raw
+        elif isinstance(raw, dict):
+            rel_path = raw.get(output_type) or list(raw.values())[0]
+
+        if not rel_path:
+            return None
+
+        # 特殊处理：从 step_outputs 记录中查找实际输出路径
+        if step_id in self.step_outputs:
+            actual_outputs = self.step_outputs[step_id]
+            if output_type in actual_outputs:
+                candidate = actual_outputs[output_type]
+                if output_type in self.MASK_TYPES and not self._is_scientific_mask(candidate):
+                    pass
+                else:
+                    return candidate
+
+        if output_type == 'water_mask':
+            if isinstance(rel_path, list):
+                for candidate in rel_path:
+                    mask_path = work_path / candidate
+                    if mask_path.exists():
+                        return str(mask_path)
+            elif rel_path:
+                mask_path = work_path / rel_path
+                if mask_path.exists():
+                    return str(mask_path)
+        elif output_type == 'reference_img':
+            if ref_img_path and Path(ref_img_path).exists():
+                return ref_img_path
+        elif output_type == 'deglint_image':
+            if isinstance(rel_path, list):
+                for candidate in rel_path:
+                    deglint_path = work_path / candidate
+                    if deglint_path.exists():
+                        return str(deglint_path)
+            elif rel_path:
+                deglint_path = work_path / rel_path
+                if deglint_path.exists():
+                    return str(deglint_path)
+            deglint_dir = work_path / "3_deglint"
+            if deglint_dir.exists():
+                for file_path in deglint_dir.glob("deglint_*.bsq"):
+                    return str(file_path)
+                for file_path in deglint_dir.glob("interpolated_*.bsq"):
+                    return str(file_path)
+        elif isinstance(rel_path, str):
+            if rel_path.endswith('/'):
+                output_path = work_path / rel_path.rstrip('/')
+                if output_path.exists() and output_path.is_dir():
+                    return str(output_path)
+            else:
+                output_path = work_path / rel_path
+                if output_path.exists():
+                    return str(output_path)
+
+        return None
+
+    def scan_work_directory_for_files(self, work_path):
+        """扫描工作目录，自动发现各步骤的输出文件
+
+        Returns:
+            discovered_outputs: dict, {step_id: {output_type: path_str}}
+        """
+        discovered_outputs = {}
+
+        subdirs = {
+            '1_water_mask': 'step1',
+            '2_Glint_Detection': 'step2',
+            '3_deglint': 'step3',
+            '5_Data_Cleaning': 'step5_clean',
+            '6_Spectral_Feature_Extraction': 'step6_feature',
+            '7_Water_Quality_Indices': 'step7_index',
+            '8_Supervised_Model_Training': 'step8_ml_train',
+            '8_Regression_Modeling': 'step8_ml_train',
+            '13_Custom_Regression': 'step13',
+            '9_ML_Prediction': 'step9_ml_predict',
+            '11_12_13_predictions/Non_Empirical_Prediction': 'step11_map',
+            '13_Custom_Regression/Custom_Regression_Prediction': 'step13',
+            '14_visualization': 'step13_report',
+            '10_geotiff_batch_rendering': 'step11_map'
+        }
+
+        for subdir, step_ids in subdirs.items():
+            subdir_path = work_path / subdir
+            if not subdir_path.exists():
+                continue
+
+            if isinstance(step_ids, str):
+                step_ids = [step_ids]
+
+            for file_path in subdir_path.rglob('*'):
+                if file_path.is_file():
+                    file_name = file_path.name.lower()
+
+                    for step_id in step_ids:
+                        if step_id not in discovered_outputs:
+                            discovered_outputs[step_id] = {}
+
+                        if 'water_mask' in file_name and step_id == 'step1':
+                            if self._is_scientific_mask(file_path):
+                                discovered_outputs[step_id]['water_mask'] = str(file_path)
+                        elif 'glint' in file_name and 'mask' in file_name and step_id == 'step2':
+                            if self._is_scientific_mask(file_path):
+                                discovered_outputs[step_id]['glint_mask'] = str(file_path)
+                        elif 'deglint' in file_name and step_id == 'step3':
+                            discovered_outputs[step_id]['deglint_image'] = str(file_path)
+                        elif 'processed_data' in file_name and step_id == 'step4_sampling':
+                            discovered_outputs[step_id]['processed_data'] = str(file_path)
+                        elif 'training_spectra' in file_name and step_id == 'step5_clean':
+                            discovered_outputs[step_id]['training_spectra'] = str(file_path)
+                        elif 'water_quality_indices' in file_name and step_id == 'step6_feature':
+                            discovered_outputs[step_id]['water_indices'] = str(file_path)
+                        elif 'sampling_spectra' in file_name and step_id == 'step4_sampling':
+                            discovered_outputs[step_id]['sampling_points'] = str(file_path)
+                        elif file_name.endswith('.csv') and step_id in ['step9_ml_predict', 'step11_map', 'step12_viz']:
+                            discovered_outputs[step_id]['predictions'] = str(file_path)
+
+        for step_id, outputs in discovered_outputs.items():
+            if step_id not in self.step_outputs:
+                self.step_outputs[step_id] = {}
+            self.step_outputs[step_id].update(outputs)
+            # ★ 发布 EventBus 事件，驱动下游面板自动填充
+            self._publish_outputs(step_id, outputs)
+
+        return discovered_outputs
+
+    def update_step_outputs(self, step_name, work_path):
+        """更新指定步骤的输出路径记录并发布 EventBus 事件。"""
+        if step_name not in self.step_default_outputs:
+            return
+
+        step_outputs = self.step_default_outputs[step_name]
+        published = {}
+
+        for output_type, relative_path in step_outputs.items():
+            if isinstance(relative_path, list):
+                for candidate in relative_path:
+                    output_path = work_path / candidate
+                    if output_path.exists():
+                        path_str = str(output_path)
+                        self.step_outputs.setdefault(step_name, {})[output_type] = path_str
+                        published[output_type] = path_str
+                        break
+            elif '*' in relative_path:
+                pattern_path = work_path / relative_path.replace('*', '*')
+                matching_files = list(pattern_path.parent.glob(pattern_path.name))
+                if matching_files:
+                    latest_file = max(matching_files, key=lambda p: p.stat().st_mtime)
+                    path_str = str(latest_file)
+                    self.step_outputs.setdefault(step_name, {})[output_type] = path_str
+                    published[output_type] = path_str
+            else:
+                output_path = work_path / relative_path
+                if output_path.exists():
+                    path_str = str(output_path)
+                    self.step_outputs.setdefault(step_name, {})[output_type] = path_str
+                    published[output_type] = path_str
+
+        if published:
+            self._publish_outputs(step_name, published)
+
+    @staticmethod
+    def prune_config_for_prediction_mode(config: dict) -> dict:
+        """Prediction-only 模式：禁用训练相关步骤，保留预测和成图步骤。
+
+        被禁用的 step dict 中统一写入 'enabled': False，
+        这些配置最终传给 PipelineRunner，Runner 会跳过它们。
+        同时，被跳过的步骤的 required_input_files 在 build_missing_items
+        中不会被检查，从而自然规避了"CSV 缺失"等训练模式下的误报。
+
+        Args:
+            config: 完整配置字典（来自 get_current_config）
+
+        Returns:
+            裁剪后的 config（深拷贝，原 config 不被修改）
+        """
+        cfg = copy.deepcopy(config)
+
+        training_steps = [
+            "step4",
+            "step5",
+            "step7",
+            "step6",
+            "step8_non_empirical_modeling",
+            "step9",
+        ]
+        for step_id in training_steps:
+            step_cfg = cfg.setdefault(step_id, {})
+            step_cfg["enabled"] = False
+
+        return cfg

src/gui/components/chart_dialogs.py

@@ -0,0 +1,471 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+图表与交互弹窗模块
+
+包含 ChartViewerDialog、ChartBrowserDialog 和 InteractiveViewerDialog 类。
+"""
+
+import numpy as np
+import pandas as pd
+from PyQt5.QtWidgets import (
+    QDialog, QVBoxLayout, QHBoxLayout, QPushButton,
+    QSizePolicy, QFileDialog, QMessageBox, QGroupBox,
+    QListWidget, QLabel, QComboBox, QCheckBox,
+)
+from PyQt5.QtCore import Qt, QAbstractTableModel
+from matplotlib.backends.backend_qt5agg import FigureCanvasQTAgg as FigureCanvas
+from matplotlib.backends.backend_qt5agg import NavigationToolbar2QT as NavigationToolbar
+from matplotlib.figure import Figure
+
+
+class ChartViewerDialog(QDialog):
+    """图表查看器对话框"""
+    def __init__(self, title="图表查看器", parent=None):
+        super().__init__(parent)
+        self.setWindowTitle(title)
+        self.resize(1000, 700)
+        self.init_ui()
+
+    def init_ui(self):
+        layout = QVBoxLayout()
+
+        self.figure = Figure(figsize=(10, 7))
+        self.canvas = FigureCanvas(self.figure)
+        self.canvas.setSizePolicy(QSizePolicy.Expanding, QSizePolicy.Expanding)
+
+        self.toolbar = NavigationToolbar(self.canvas, self)
+
+        layout.addWidget(self.toolbar)
+        layout.addWidget(self.canvas)
+
+        btn_layout = QHBoxLayout()
+
+        self.save_btn = QPushButton("保存图表")
+        self.save_btn.clicked.connect(self.save_chart)
+        btn_layout.addWidget(self.save_btn)
+
+        btn_layout.addStretch()
+
+        self.close_btn = QPushButton("关闭")
+        self.close_btn.clicked.connect(self.close)
+        btn_layout.addWidget(self.close_btn)
+
+        layout.addLayout(btn_layout)
+        self.setLayout(layout)
+
+    def display_image(self, image_path):
+        """显示图片"""
+        self.figure.clear()
+        ax = self.figure.add_subplot(111)
+
+        try:
+            import matplotlib.image as mpimg
+            img = mpimg.imread(image_path)
+            ax.imshow(img)
+            ax.axis('off')
+            self.figure.tight_layout()
+            self.canvas.draw()
+            self.current_image_path = image_path
+        except Exception as e:
+            ax.text(0.5, 0.5, f'加载图片失败:\n{str(e)}',
+                   ha='center', va='center', transform=ax.transAxes)
+            self.canvas.draw()
+
+    def display_custom_plot(self, plot_func):
+        """显示自定义绘图函数"""
+        self.figure.clear()
+        try:
+            plot_func(self.figure)
+            self.canvas.draw()
+        except Exception as e:
+            ax = self.figure.add_subplot(111)
+            ax.text(0.5, 0.5, f'绘图失败:\n{str(e)}',
+                   ha='center', va='center', transform=ax.transAxes)
+            self.canvas.draw()
+
+    def save_chart(self):
+        """保存图表"""
+        file_path, _ = QFileDialog.getSaveFileName(
+            self, "保存图表", "",
+            "PNG图片 (*.png);;JPG图片 (*.jpg);;PDF文件 (*.pdf);;所有文件 (*.*)"
+        )
+        if file_path:
+            try:
+                self.figure.savefig(file_path, dpi=300, bbox_inches='tight')
+                QMessageBox.information(self, "成功", f"图表已保存到:\n{file_path}")
+            except Exception as e:
+                QMessageBox.critical(self, "错误", f"保存失败:\n{str(e)}")
+
+
+class ChartBrowserDialog(QDialog):
+    """图表浏览器对话框"""
+    def __init__(self, chart_files, parent=None):
+        super().__init__(parent)
+        self.chart_files = sorted(chart_files, key=lambda x: x.stat().st_mtime, reverse=True)
+        self.current_index = 0
+        self.setWindowTitle("图表浏览器")
+        self.resize(1200, 800)
+        self.init_ui()
+        self.show_chart(0)
+
+    def init_ui(self):
+        layout = QVBoxLayout()
+
+        list_group = QGroupBox(f"图表列表 (共 {len(self.chart_files)} 个)")
+        list_layout = QHBoxLayout()
+
+        self.chart_list = QListWidget()
+        self.chart_list.setMaximumHeight(150)
+        for chart_file in self.chart_files:
+            self.chart_list.addItem(chart_file.name)
+        self.chart_list.currentRowChanged.connect(self.show_chart)
+
+        list_layout.addWidget(self.chart_list)
+        list_group.setLayout(list_layout)
+        layout.addWidget(list_group)
+
+        self.figure = Figure(figsize=(12, 8))
+        self.canvas = FigureCanvas(self.figure)
+        self.canvas.setSizePolicy(QSizePolicy.Expanding, QSizePolicy.Expanding)
+
+        self.toolbar = NavigationToolbar(self.canvas, self)
+        layout.addWidget(self.toolbar)
+        layout.addWidget(self.canvas, 1)
+
+        btn_layout = QHBoxLayout()
+
+        self.prev_btn = QPushButton("◀ 上一个")
+        self.prev_btn.clicked.connect(self.prev_chart)
+        btn_layout.addWidget(self.prev_btn)
+
+        self.next_btn = QPushButton("下一个 >")
+        self.next_btn.clicked.connect(self.next_chart)
+        btn_layout.addWidget(self.next_btn)
+
+        btn_layout.addStretch()
+
+        self.save_btn = QPushButton("💾 保存当前图表")
+        self.save_btn.clicked.connect(self.save_current_chart)
+        btn_layout.addWidget(self.save_btn)
+
+        self.close_btn = QPushButton("关闭")
+        self.close_btn.clicked.connect(self.close)
+        btn_layout.addWidget(self.close_btn)
+
+        layout.addLayout(btn_layout)
+        self.setLayout(layout)
+
+    def show_chart(self, index):
+        """显示指定索引的图表"""
+        if 0 <= index < len(self.chart_files):
+            self.current_index = index
+            self.chart_list.setCurrentRow(index)
+
+            chart_file = self.chart_files[index]
+            self.figure.clear()
+            ax = self.figure.add_subplot(111)
+
+            try:
+                import matplotlib.image as mpimg
+                img = mpimg.imread(str(chart_file))
+                ax.imshow(img)
+                ax.axis('off')
+                ax.set_title(chart_file.name, fontsize=12, pad=10)
+                self.figure.tight_layout()
+                self.canvas.draw()
+            except Exception as e:
+                ax.text(0.5, 0.5, f'加载图片失败:\n{str(e)}',
+                       ha='center', va='center', transform=ax.transAxes)
+                self.canvas.draw()
+
+            self.prev_btn.setEnabled(index > 0)
+            self.next_btn.setEnabled(index < len(self.chart_files) - 1)
+
+    def prev_chart(self):
+        """上一个图表"""
+        if self.current_index > 0:
+            self.show_chart(self.current_index - 1)
+
+    def next_chart(self):
+        """下一个图表"""
+        if self.current_index < len(self.chart_files) - 1:
+            self.show_chart(self.current_index + 1)
+
+    def save_current_chart(self):
+        """保存当前图表"""
+        if 0 <= self.current_index < len(self.chart_files):
+            current_file = self.chart_files[self.current_index]
+            file_path, _ = QFileDialog.getSaveFileName(
+                self, "保存图表", current_file.name,
+                "PNG图片 (*.png);;JPG图片 (*.jpg);;所有文件 (*.*)"
+            )
+            if file_path:
+                try:
+                    import shutil
+                    shutil.copy(str(current_file), file_path)
+                    QMessageBox.information(self, "成功", f"图表已保存到:\n{file_path}")
+                except Exception as e:
+                    QMessageBox.critical(self, "错误", f"保存失败:\n{str(e)}")
+
+
+class InteractiveViewerDialog(QDialog):
+    """交互式影像预览对话框：显示影像、参考点散点图、点击查询坐标/值"""
+
+    def __init__(self, parent, img_path, ref_csv=None):
+        super().__init__(parent)
+        self.img_path = img_path
+        self.ref_csv = ref_csv
+        self.geotransform = None
+        self.fig = None
+        self.canvas = None
+        self.ax = None
+        self.status_label = None
+        self.init_ui()
+
+    def init_ui(self):
+        self.setWindowTitle("👁️ 交互式影像预览")
+        self.setMinimumSize(900, 700)
+
+        layout = QVBoxLayout()
+
+        toolbar = QHBoxLayout()
+        self.band_combo = QComboBox()
+        self.band_combo.currentIndexChanged.connect(self.on_band_changed)
+        toolbar.addWidget(QLabel("显示波段:"))
+        toolbar.addWidget(self.band_combo)
+
+        self.gray_check = QCheckBox("灰度显示")
+        self.gray_check.stateChanged.connect(self.on_band_changed)
+        toolbar.addWidget(self.gray_check)
+        toolbar.addStretch()
+        layout.addLayout(toolbar)
+
+        try:
+            from matplotlib.backends.backend_qt5agg import FigureCanvasQTAgg as FigureCanvas
+            from matplotlib.figure import Figure
+            import matplotlib
+            matplotlib.use('Qt5Agg')
+
+            self.fig = Figure(figsize=(10, 8))
+            self.canvas = FigureCanvas(self.fig)
+            self.ax = self.fig.add_subplot(111)
+            self.fig.tight_layout()
+            layout.addWidget(self.canvas)
+
+            self.load_and_display()
+
+        except ImportError as e:
+            layout.addWidget(QLabel(f"Matplotlib 未安装: {e}"))
+
+        self.status_label = QLabel("点击影像查看像素坐标和经纬度")
+        self.status_label.setStyleSheet("background:#f0f0f0;padding:4px;font-size:12px;")
+        self.status_label.setWordWrap(True)
+        layout.addWidget(self.status_label)
+
+        close_btn = QPushButton("关闭")
+        close_btn.clicked.connect(self.close)
+        layout.addWidget(close_btn)
+
+        self.setLayout(layout)
+
+    def load_and_display(self):
+        """加载影像并显示"""
+        from osgeo import gdal
+
+        dataset = gdal.Open(self.img_path)
+        if dataset is None:
+            self.status_label.setText(f"无法打开影像: {self.img_path}")
+            return
+
+        self.geotransform = dataset.GetGeoTransform()
+        self.projection = dataset.GetProjection()
+        n_bands = dataset.RasterCount
+        self.height = dataset.RasterYSize
+        self.width = dataset.RasterXSize
+
+        self.band_combo.clear()
+        if n_bands >= 3:
+            for i in range(1, n_bands + 1):
+                self.band_combo.addItem(f"RGB (B{i-0}, G{i-1}, R{i-2})" if i >= 3 else f"波段 {i}", i)
+            self.band_combo.addItem(f"单波段 (B1)", 0)
+        else:
+            for i in range(1, n_bands + 1):
+                self.band_combo.addItem(f"波段 {i}", i - 1)
+        self.band_combo.setCurrentIndex(0)
+
+        self.dataset = dataset
+        self.display_band(0, is_gray=False)
+        self.load_ref_points()
+
+    def display_band(self, band_idx, is_gray=False):
+        """显示指定波段组合"""
+        from osgeo import gdal
+        import numpy as np
+
+        dataset = self.dataset
+        self.ax.clear()
+
+        if is_gray or (self.band_combo.currentData() == 0 and dataset.RasterCount == 1):
+            band = dataset.GetRasterBand(1 if band_idx == 0 else band_idx + 1)
+            data = band.ReadAsArray()
+            data = np.nan_to_num(data, nan=0.0)
+            self.ax.imshow(data, cmap='gray')
+            self.ax.set_title(f"波段 {band_idx + 1} (灰度)")
+        else:
+            n = min(3, dataset.RasterCount)
+            bands_data = []
+            for i in range(n):
+                b = dataset.GetRasterBand(i + 1)
+                bd = b.ReadAsArray()
+                bd = np.nan_to_num(bd, nan=0.0)
+                bands_data.append(bd)
+            rgb = np.dstack(bands_data)
+
+            for i in range(rgb.shape[2]):
+                p2, p98 = np.percentile(rgb[:, :, i], [2, 98])
+                if p98 > p2:
+                    rgb[:, :, i] = np.clip((rgb[:, :, i] - p2) / (p98 - p2), 0, 1)
+                else:
+                    rgb[:, :, i] = np.clip(rgb[:, :, i] / (p98 + 1e-6), 0, 1)
+
+            self.ax.imshow(rgb)
+            self.ax.set_title(f"RGB 显示")
+
+        self.ax.set_xlabel("列 (Column)")
+        self.ax.set_ylabel("行 (Row)")
+        self.fig.tight_layout()
+        self.canvas.draw()
+
+        self.cid = self.canvas.mpl_connect('button_press_event', self.on_click)
+
+    def on_band_changed(self):
+        """波段选择变化时更新显示"""
+        if not hasattr(self, 'dataset'):
+            return
+        is_gray = self.gray_check.isChecked()
+        band_data = self.band_combo.currentData()
+        self.display_band(band_data if band_data != 0 else 0, is_gray=is_gray)
+
+    def load_ref_points(self):
+        """加载并显示参考点"""
+        import os
+        if not self.ref_csv or not os.path.isfile(self.ref_csv):
+            return
+
+        try:
+            import csv
+            lon_list, lat_list = [], []
+            with open(self.ref_csv, 'r', encoding='utf-8-sig') as f:
+                reader = csv.DictReader(f)
+                for row in reader:
+                    try:
+                        lon = float(row.get('Lon', row.get('lon', row.get('LON', 0))))
+                        lat = float(row.get('Lat', row.get('lat', row.get('LAT', 0))))
+                        if lon and lat:
+                            lon_list.append(lon)
+                            lat_list.append(lat)
+                    except (ValueError, TypeError):
+                        continue
+
+            if not lon_list:
+                return
+
+            px_list, py_list = [], []
+            gt = self.geotransform
+            if gt and (gt[1] != 0 or gt[5] != 0):
+                for lon, lat in zip(lon_list, lat_list):
+                    px = (lon - gt[0]) / gt[1]
+                    py = (lat - gt[3]) / gt[5]
+                    if 0 <= px < self.width and 0 <= py < self.height:
+                        px_list.append(px)
+                        py_list.append(py)
+
+            if px_list:
+                self.ax.scatter(px_list, py_list, c='red', s=40, marker='o',
+                               edgecolors='white', linewidths=0.8, zorder=5, alpha=0.9,
+                               label=f'参考点 ({len(px_list)}个)')
+                self.ax.legend(loc='upper right', fontsize=9)
+                self.fig.tight_layout()
+                self.canvas.draw()
+                self.status_label.setText(
+                    f"已加载 {len(px_list)} 个参考点（仅显示在影像范围内的点）"
+                )
+        except Exception as e:
+            self.status_label.setText(f"加载参考点失败: {e}")
+
+    def pixel_to_geo(self, px, py):
+        """像素坐标转经纬度"""
+        gt = self.geotransform
+        if gt is None:
+            return None, None
+        lon = gt[0] + px * gt[1] + py * gt[2]
+        lat = gt[3] + px * gt[4] + py * gt[5]
+        return lon, lat
+
+    def on_click(self, event):
+        """鼠标点击事件"""
+        if event.inaxes != self.ax or event.xdata is None or event.ydata is None:
+            return
+
+        px, py = int(round(event.xdata)), int(round(event.ydata))
+        if not (0 <= px < self.width and 0 <= py < self.height):
+            return
+
+        from osgeo import gdal
+        import numpy as np
+        dataset = self.dataset
+        n_bands = dataset.RasterCount
+        vals = []
+        for b in range(1, n_bands + 1):
+            val = dataset.GetRasterBand(b).ReadAsArray()[py, px]
+            vals.append(f"{val:.4f}" if isinstance(val, float) else str(val))
+
+        lon, lat = self.pixel_to_geo(px, py)
+        geo_str = f"Lon={lon:.6f}, Lat={lat:.6f}" if lon is not None else "无地理参考"
+
+        self.status_label.setText(
+            f"像素: (行={py}, 列={px}) | {geo_str} | "
+            f"波段值: {' | '.join(vals[:5])}" +
+            (f" ... ({n_bands}波段的更多信息)" if n_bands > 5 else "")
+        )
+
+
+class PandasTableModel(QAbstractTableModel):
+    """支持DataFrame的表格模型"""
+    def __init__(self, data_frame: pd.DataFrame):
+        super().__init__()
+        self._data = data_frame.copy()
+        if self._data.empty:
+            self._data = pd.DataFrame()
+        self._data.fillna("", inplace=True)
+        self._columns = [str(col) for col in self._data.columns]
+
+    def rowCount(self, parent=None):
+        return len(self._data)
+
+    def columnCount(self, parent=None):
+        return len(self._columns)
+
+    def data(self, index, role=Qt.DisplayRole):
+        if not index.isValid() or role != Qt.DisplayRole:
+            return None
+
+        value = self._data.iat[index.row(), index.column()]
+        if pd.isna(value):
+            return ""
+        return str(value)
+
+    def headerData(self, section, orientation, role=Qt.DisplayRole):
+        if role != Qt.DisplayRole:
+            return None
+        if orientation == Qt.Horizontal:
+            if section < len(self._columns):
+                return self._columns[section]
+            return str(section)
+        return str(section + 1)
+
+    def flags(self, index):
+        if not index.isValid():
+            return Qt.NoItemFlags
+        return Qt.ItemIsEnabled | Qt.ItemIsSelectable

src/gui/components/data_models.py

@@ -0,0 +1,50 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+数据模型模块
+
+包含 PandasTableModel 等数据模型类。
+"""
+
+import pandas as pd
+from PyQt5.QtCore import Qt, QAbstractTableModel
+
+
+class PandasTableModel(QAbstractTableModel):
+    """支持DataFrame的表格模型"""
+    def __init__(self, data_frame: pd.DataFrame):
+        super().__init__()
+        self._data = data_frame.copy()
+        if self._data.empty:
+            self._data = pd.DataFrame()
+        self._data.fillna("", inplace=True)
+        self._columns = [str(col) for col in self._data.columns]
+
+    def rowCount(self, parent=None):
+        return len(self._data)
+
+    def columnCount(self, parent=None):
+        return len(self._columns)
+
+    def data(self, index, role=Qt.DisplayRole):
+        if not index.isValid() or role != Qt.DisplayRole:
+            return None
+
+        value = self._data.iat[index.row(), index.column()]
+        if pd.isna(value):
+            return ""
+        return str(value)
+
+    def headerData(self, section, orientation, role=Qt.DisplayRole):
+        if role != Qt.DisplayRole:
+            return None
+        if orientation == Qt.Horizontal:
+            if section < len(self._columns):
+                return self._columns[section]
+            return str(section)
+        return str(section + 1)
+
+    def flags(self, index):
+        if not index.isValid():
+            return Qt.NoItemFlags
+        return Qt.ItemIsEnabled | Qt.ItemIsSelectable

src/gui/components/image_viewer_components.py

@@ -0,0 +1,374 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+图像查看器组件模块
+
+包含 ImageCategoryTree 和 ImageViewerWidget 类。
+"""
+
+import os
+from pathlib import Path
+from typing import Optional, List
+
+from PyQt5.QtWidgets import (
+    QWidget, QVBoxLayout, QHBoxLayout, QPushButton,
+    QFrame, QScrollArea, QLabel, QFileDialog, QMessageBox,
+    QTreeWidget, QTreeWidgetItem,
+)
+from PyQt5.QtCore import Qt, QTimer
+from PyQt5.QtGui import QPixmap
+
+
+class ImageCategoryTree(QTreeWidget):
+    """图像分类目录树 - 按类别组织图像文件"""
+
+    # 图像类别定义：(类别名称, 关键词列表, 图标)
+    CATEGORIES = [
+        ("模型评估", ["scatter", "regression", "validation", "r2", "rmse"], "📊"),
+        ("光谱分析", ["spectrum", "spectral", "band", "wavelength"], "📈"),
+        ("统计图表", ["boxplot", "histogram", "heatmap", "statistics", "stats"], "📉"),
+        ("处理结果", ["mask", "glint", "deglint", "preview", "overlay", "water_mask"], "🖼️"),
+        ("含量分布图", [], "📁"),
+    ]
+
+    def __init__(self, parent=None):
+        super().__init__(parent)
+        self.setHeaderLabel("图像目录")
+        self.setMaximumWidth(300)
+        self.setMinimumWidth(250)
+        self.setup_categories()
+        self.setStyleSheet("""
+            QTreeWidget {
+                border: 1px solid #ddd;
+                border-radius: 5px;
+                background-color: #f8f9fa;
+            }
+            QTreeWidget::item {
+                padding: 5px;
+                border-radius: 3px;
+            }
+            QTreeWidget::item:selected {
+                background-color: #0078D4;
+                color: white;
+            }
+            QTreeWidget::item:hover {
+                background-color: #e3f2fd;
+            }
+        """)
+
+    def setup_categories(self):
+        """初始化类别节点"""
+        self.category_items = {}
+        for category_name, keywords, icon in self.CATEGORIES:
+            item = QTreeWidgetItem(self)
+            item.setText(0, f"{icon} {category_name}")
+            item.setData(0, Qt.UserRole, {"type": "category", "keywords": keywords, "name": category_name})
+            item.setExpanded(True)
+            self.category_items[category_name] = item
+
+    def clear_all_images(self):
+        """清除所有图像项"""
+        for category_item in self.category_items.values():
+            # 删除所有子项
+            while category_item.childCount() > 0:
+                category_item.removeChild(category_item.child(0))
+
+    def add_image(self, file_path: Path, display_name: str = None):
+        """添加图像到对应的类别"""
+        if display_name is None:
+            display_name = file_path.stem
+
+        # 根据文件名关键词确定类别
+        category = self._determine_category(file_path.name)
+        category_item = self.category_items.get(category, self.category_items["含量分布图"])
+
+        # 创建图像项
+        image_item = QTreeWidgetItem(category_item)
+        image_item.setText(0, f"  └─ {display_name}")
+        image_item.setData(0, Qt.UserRole, {"type": "image", "path": str(file_path)})
+        image_item.setToolTip(0, str(file_path))
+
+        return image_item
+
+    def _determine_category(self, filename: str) -> str:
+        """根据文件名确定类别"""
+        filename_lower = filename.lower()
+
+        for category_name, keywords, _ in self.CATEGORIES:
+            if any(keyword in filename_lower for keyword in keywords):
+                return category_name
+
+        return "含量分布图"
+
+    def scan_directory(self, work_dir: str):
+        """扫描目录中的所有图像文件"""
+        self.clear_all_images()
+
+        work_path = Path(work_dir)
+        if not work_path.exists():
+            return
+
+        # 查找所有图像文件：14_visualization 为主，同时扫描步骤产出目录（如 1_water_mask 下的预览/叠置图）
+        image_extensions = ['*.png', '*.jpg', '*.jpeg', '*.tif', '*.tiff', '*.bmp']
+        scan_roots: List[Path] = []
+        _viz = work_path / "14_visualization"
+        if _viz.is_dir():
+            scan_roots.append(_viz)
+        _wm = work_path / "1_water_mask"
+        if _wm.is_dir():
+            scan_roots.append(_wm)
+        if not scan_roots:
+            scan_roots.append(work_path)
+
+        seen_norm: set = set()
+        image_files: List[Path] = []
+        for root in scan_roots:
+            for ext in image_extensions:
+                for p in root.glob(f"**/{ext}"):
+                    key = os.path.normcase(os.path.normpath(str(p.resolve())))
+                    if key in seen_norm:
+                        continue
+                    seen_norm.add(key)
+                    image_files.append(p)
+
+        # 添加图像到树
+        for img_file in sorted(image_files):
+            # 跳过缩略图和临时文件
+            if img_file.name.startswith('.') or 'thumb' in img_file.name.lower():
+                continue
+            self.add_image(img_file)
+
+        # 更新类别项文本显示数量
+        for category_name, item in self.category_items.items():
+            count = item.childCount()
+            if count > 0:
+                for cat_name, _, icon in self.CATEGORIES:
+                    if cat_name == category_name:
+                        item.setText(0, f"{icon} {category_name} ({count})")
+                        break
+
+    def get_selected_image_path(self) -> Optional[str]:
+        """获取当前选中的图像路径"""
+        selected_item = self.currentItem()
+        if not selected_item:
+            return None
+
+        data = selected_item.data(0, Qt.UserRole)
+        if data and data.get("type") == "image":
+            return data.get("path")
+        return None
+
+
+class ImageViewerWidget(QWidget):
+    """图像查看器组件 - 支持缩放、平移"""
+
+    def __init__(self, parent=None):
+        super().__init__(parent)
+        self.current_image_path = None
+        self.scale_factor = 1.0
+        self._update_timer = QTimer()  # 防抖定时器
+        self._update_timer.setSingleShot(True)
+        self._update_timer.timeout.connect(self._do_update_display)
+        self._pending_scale = None  # 待更新的缩放比例
+        self.setup_ui()
+
+    def setup_ui(self):
+        layout = QVBoxLayout()
+        layout.setContentsMargins(0, 0, 0, 0)
+
+        # 工具栏
+        toolbar = QHBoxLayout()
+
+        self.refresh_btn = QPushButton("🔄 刷新目录")
+        self.refresh_btn.setToolTip("重新扫描工作目录中的图像文件")
+        toolbar.addWidget(self.refresh_btn)
+
+        # 添加分隔线
+        separator = QFrame()
+        separator.setFrameShape(QFrame.VLine)
+        separator.setFrameShadow(QFrame.Sunken)
+        toolbar.addWidget(separator)
+
+        self.zoom_in_btn = QPushButton("🔍+")
+        self.zoom_in_btn.setToolTip("放大")
+        self.zoom_in_btn.setMaximumWidth(50)
+        toolbar.addWidget(self.zoom_in_btn)
+
+        self.zoom_out_btn = QPushButton("🔍-")
+        self.zoom_out_btn.setToolTip("缩小")
+        self.zoom_out_btn.setMaximumWidth(50)
+        toolbar.addWidget(self.zoom_out_btn)
+
+        self.fit_btn = QPushButton("⬜ 适应窗口")
+        self.fit_btn.setToolTip("适应窗口大小")
+        toolbar.addWidget(self.fit_btn)
+
+        self.original_btn = QPushButton("1:1 原始大小")
+        self.original_btn.setToolTip("原始大小")
+        toolbar.addWidget(self.original_btn)
+
+        toolbar.addStretch()
+
+        self.save_btn = QPushButton("💾 保存")
+        self.save_btn.setToolTip("保存当前图像")
+        toolbar.addWidget(self.save_btn)
+
+        layout.addLayout(toolbar)
+
+        # 图像显示区域 - 使用 QLabel + QScrollArea
+        self.scroll_area = QScrollArea()
+        self.scroll_area.setWidgetResizable(True)
+        self.scroll_area.setStyleSheet("background-color: white;")
+
+        self.image_label = QLabel()
+        self.image_label.setAlignment(Qt.AlignCenter)
+        self.image_label.setStyleSheet("background-color:  white;")
+
+        self.scroll_area.setWidget(self.image_label)
+        layout.addWidget(self.scroll_area, 1)
+
+        # 状态栏
+        status_layout = QHBoxLayout()
+        self.status_label = QLabel("就绪")
+        self.status_label.setStyleSheet("color: #666; font-size: 11px;")
+        status_layout.addWidget(self.status_label)
+        status_layout.addStretch()
+        layout.addLayout(status_layout)
+
+        self.setLayout(layout)
+
+        # 连接信号
+        self.zoom_in_btn.clicked.connect(self.zoom_in)
+        self.zoom_out_btn.clicked.connect(self.zoom_out)
+        self.fit_btn.clicked.connect(self.fit_to_window)
+        self.original_btn.clicked.connect(self.original_size)
+        self.save_btn.clicked.connect(self.save_image)
+
+    def load_image(self, image_path: str):
+        """加载并显示图像"""
+        if not image_path or not Path(image_path).exists():
+            self.image_label.setText("图像不存在")
+            self.status_label.setText("图像加载失败")
+            return
+
+        self.current_image_path = image_path
+        self.scale_factor = 1.0
+
+        # 加载图像
+        pixmap = QPixmap(image_path)
+        if pixmap.isNull():
+            self.image_label.setText("无法加载图像")
+            self.status_label.setText("图像格式不支持")
+            return
+
+        self.original_pixmap = pixmap
+
+        # 默认适应窗口显示
+        self.fit_to_window()
+
+        # 更新状态
+        file_info = Path(image_path).stat()
+        size_mb = file_info.st_size / (1024 * 1024)
+        self.status_label.setText(f"{pixmap.width()}x{pixmap.height()} | {size_mb:.2f} MB | {Path(image_path).name} | 适应窗口")
+
+    def update_image_display(self):
+        """更新图像显示 - 使用防抖避免频繁重绘卡顿"""
+        # 取消之前的待执行更新，重新计时
+        self._update_timer.stop()
+        self._pending_scale = self.scale_factor
+        self._update_timer.start(50)  # 50ms后执行实际更新
+
+    def _do_update_display(self):
+        """实际执行图像更新"""
+        if not hasattr(self, 'original_pixmap') or self.original_pixmap.isNull():
+            return
+
+        if self._pending_scale is None:
+            return
+
+        # 根据缩放比例选择变换模式：大幅度缩放用Fast模式提升性能
+        if self._pending_scale > 2.0 or self._pending_scale < 0.5:
+            transform = Qt.FastTransformation
+        else:
+            transform = Qt.SmoothTransformation
+
+        scaled_pixmap = self.original_pixmap.scaled(
+            int(self.original_pixmap.width() * self._pending_scale),
+            int(self.original_pixmap.height() * self._pending_scale),
+            Qt.KeepAspectRatio,
+            transform
+        )
+        self.image_label.setPixmap(scaled_pixmap)
+        self._pending_scale = None
+
+    def wheelEvent(self, event):
+        """鼠标滚轮缩放 - 实时响应"""
+        delta = event.angleDelta().y()
+
+        if delta > 0:
+            # 向上滚动 - 放大
+            if self.scale_factor < 5.0:
+                self.scale_factor = min(self.scale_factor * 1.1, 5.0)
+                self.update_image_display()
+        else:
+            # 向下滚动 - 缩小
+            if self.scale_factor > 0.1:
+                self.scale_factor = max(self.scale_factor / 1.1, 0.1)
+                self.update_image_display()
+
+        event.accept()
+
+    def zoom_in(self):
+        """放大"""
+        if self.scale_factor < 5.0:
+            self.scale_factor = min(self.scale_factor * 1.25, 5.0)
+            self.update_image_display()
+
+    def zoom_out(self):
+        """缩小"""
+        if self.scale_factor > 0.1:
+            self.scale_factor = max(self.scale_factor / 1.25, 0.1)
+            self.update_image_display()
+
+    def fit_to_window(self):
+        """适应窗口"""
+        if not hasattr(self, 'original_pixmap') or self.original_pixmap.isNull():
+            return
+
+        # 计算适应窗口的缩放比例
+        view_size = self.scroll_area.viewport().size()
+        img_size = self.original_pixmap.size()
+
+        scale_w = view_size.width() / img_size.width()
+        scale_h = view_size.height() / img_size.height()
+
+        # 记录适应前的比例（用于后续恢复参考）
+        self._fit_scale = min(scale_w, scale_h)
+        self.scale_factor = self._fit_scale
+
+        self.update_image_display()
+        self.status_label.setText(f"适应窗口 | 缩放: {self.scale_factor:.1%}")
+
+    def original_size(self):
+        """原始大小"""
+        self.scale_factor = 1.0
+        self._fit_scale = None  # 清除适应记录
+        self.update_image_display()
+        self.status_label.setText("原始大小 | 缩放: 100%")
+
+    def save_image(self):
+        """保存图像"""
+        if not self.current_image_path:
+            return
+
+        file_path, _ = QFileDialog.getSaveFileName(
+            self, "保存图像", Path(self.current_image_path).name,
+            "PNG图片 (*.png);;JPG图片 (*.jpg);;所有文件 (*.*)"
+        )
+
+        if file_path:
+            try:
+                import shutil
+                shutil.copy(self.current_image_path, file_path)
+            except Exception as e:
+                QMessageBox.critical(self, "错误", f"保存失败: {e}")

src/gui/components/image_widgets.py

@@ -0,0 +1,351 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+图像浏览组件模块
+
+包含 ImageCategoryTree 和 ImageViewerWidget 类。
+"""
+
+import os
+from pathlib import Path
+from typing import List, Optional
+
+from PyQt5.QtWidgets import (
+    QTreeWidget, QTreeWidgetItem, QWidget, QVBoxLayout, QHBoxLayout,
+    QPushButton, QLabel, QScrollArea, QFrame, QGroupBox,
+    QFileDialog, QMessageBox,
+)
+from PyQt5.QtCore import Qt, QTimer
+from PyQt5.QtGui import QPixmap
+
+
+class ImageCategoryTree(QTreeWidget):
+    """图像分类目录树 - 按类别组织图像文件"""
+
+    CATEGORIES = [
+        ("模型评估", ["scatter", "regression", "validation", "r2", "rmse"], "📊"),
+        ("光谱分析", ["spectrum", "spectral", "band", "wavelength"], "📈"),
+        ("统计图表", ["boxplot", "histogram", "heatmap", "statistics", "stats"], "📉"),
+        ("处理结果", ["mask", "glint", "deglint", "preview", "overlay", "water_mask"], "🖼️"),
+        ("含量分布图", [], "📁"),
+    ]
+
+    def __init__(self, parent=None):
+        super().__init__(parent)
+        self.setHeaderLabel("图像目录")
+        self.setMaximumWidth(300)
+        self.setMinimumWidth(250)
+        self.setup_categories()
+        self.setStyleSheet("""
+            QTreeWidget {
+                border: 1px solid #ddd;
+                border-radius: 5px;
+                background-color: #f8f9fa;
+            }
+            QTreeWidget::item {
+                padding: 5px;
+                border-radius: 3px;
+            }
+            QTreeWidget::item:selected {
+                background-color: #0078D4;
+                color: white;
+            }
+            QTreeWidget::item:hover {
+                background-color: #e3f2fd;
+            }
+        """)
+
+    def setup_categories(self):
+        """初始化类别节点"""
+        self.category_items = {}
+        for category_name, keywords, icon in self.CATEGORIES:
+            item = QTreeWidgetItem(self)
+            item.setText(0, f"{icon} {category_name}")
+            item.setData(0, Qt.UserRole, {"type": "category", "keywords": keywords, "name": category_name})
+            item.setExpanded(True)
+            self.category_items[category_name] = item
+
+    def clear_all_images(self):
+        """清除所有图像项"""
+        for category_item in self.category_items.values():
+            while category_item.childCount() > 0:
+                category_item.removeChild(category_item.child(0))
+
+    def add_image(self, file_path: Path, display_name: str = None):
+        """添加图像到对应的类别"""
+        if display_name is None:
+            display_name = file_path.stem
+
+        category = self._determine_category(file_path.name)
+        category_item = self.category_items.get(category, self.category_items["含量分布图"])
+
+        image_item = QTreeWidgetItem(category_item)
+        image_item.setText(0, f"  └─ {display_name}")
+        image_item.setData(0, Qt.UserRole, {"type": "image", "path": str(file_path)})
+        image_item.setToolTip(0, str(file_path))
+
+        return image_item
+
+    def _determine_category(self, filename: str) -> str:
+        """根据文件名确定类别"""
+        filename_lower = filename.lower()
+
+        for category_name, keywords, _ in self.CATEGORIES:
+            if any(keyword in filename_lower for keyword in keywords):
+                return category_name
+
+        return "含量分布图"
+
+    def scan_directory(self, work_dir: str):
+        """扫描目录中的所有图像文件"""
+        self.clear_all_images()
+
+        work_path = Path(work_dir)
+        if not work_path.exists():
+            return
+
+        image_extensions = ['*.png', '*.jpg', '*.jpeg', '*.tif', '*.tiff', '*.bmp']
+        scan_roots: List[Path] = []
+        _viz = work_path / "14_visualization"
+        if _viz.is_dir():
+            scan_roots.append(_viz)
+        _wm = work_path / "1_water_mask"
+        if _wm.is_dir():
+            scan_roots.append(_wm)
+        if not scan_roots:
+            scan_roots.append(work_path)
+
+        seen_norm: set = set()
+        image_files: List[Path] = []
+        for root in scan_roots:
+            for ext in image_extensions:
+                for p in root.glob(f"**/{ext}"):
+                    key = os.path.normcase(os.path.normpath(str(p.resolve())))
+                    if key in seen_norm:
+                        continue
+                    seen_norm.add(key)
+                    image_files.append(p)
+
+        for img_file in sorted(image_files):
+            if img_file.name.startswith('.') or 'thumb' in img_file.name.lower():
+                continue
+            self.add_image(img_file)
+
+        for category_name, item in self.category_items.items():
+            count = item.childCount()
+            if count > 0:
+                for cat_name, _, icon in self.CATEGORIES:
+                    if cat_name == category_name:
+                        item.setText(0, f"{icon} {category_name} ({count})")
+                        break
+
+    def get_selected_image_path(self) -> Optional[str]:
+        """获取当前选中的图像路径"""
+        selected_item = self.currentItem()
+        if not selected_item:
+            return None
+
+        data = selected_item.data(0, Qt.UserRole)
+        if data and data.get("type") == "image":
+            return data.get("path")
+        return None
+
+
+class ImageViewerWidget(QWidget):
+    """图像查看器组件 - 支持缩放、平移"""
+
+    def __init__(self, parent=None):
+        super().__init__(parent)
+        self.current_image_path = None
+        self.scale_factor = 1.0
+        self._update_timer = QTimer()
+        self._update_timer.setSingleShot(True)
+        self._update_timer.timeout.connect(self._do_update_display)
+        self._pending_scale = None
+        self.setup_ui()
+
+    def setup_ui(self):
+        layout = QVBoxLayout()
+        layout.setContentsMargins(0, 0, 0, 0)
+
+        toolbar = QHBoxLayout()
+
+        self.refresh_btn = QPushButton("🔄 刷新目录")
+        self.refresh_btn.setToolTip("重新扫描工作目录中的图像文件")
+        toolbar.addWidget(self.refresh_btn)
+
+        separator = QFrame()
+        separator.setFrameShape(QFrame.VLine)
+        separator.setFrameShadow(QFrame.Sunken)
+        toolbar.addWidget(separator)
+
+        self.zoom_in_btn = QPushButton("🔍+")
+        self.zoom_in_btn.setToolTip("放大")
+        self.zoom_in_btn.setMaximumWidth(50)
+        toolbar.addWidget(self.zoom_in_btn)
+
+        self.zoom_out_btn = QPushButton("🔍-")
+        self.zoom_out_btn.setToolTip("缩小")
+        self.zoom_out_btn.setMaximumWidth(50)
+        toolbar.addWidget(self.zoom_out_btn)
+
+        self.fit_btn = QPushButton("⬜ 适应窗口")
+        self.fit_btn.setToolTip("适应窗口大小")
+        toolbar.addWidget(self.fit_btn)
+
+        self.original_btn = QPushButton("1:1 原始大小")
+        self.original_btn.setToolTip("原始大小")
+        toolbar.addWidget(self.original_btn)
+
+        toolbar.addStretch()
+
+        self.save_btn = QPushButton("💾 保存")
+        self.save_btn.setToolTip("保存当前图像")
+        toolbar.addWidget(self.save_btn)
+
+        layout.addLayout(toolbar)
+
+        self.scroll_area = QScrollArea()
+        self.scroll_area.setWidgetResizable(True)
+        self.scroll_area.setStyleSheet("background-color: white;")
+
+        self.image_label = QLabel()
+        self.image_label.setAlignment(Qt.AlignCenter)
+        self.image_label.setStyleSheet("background-color:  white;")
+
+        self.scroll_area.setWidget(self.image_label)
+        layout.addWidget(self.scroll_area, 1)
+
+        status_layout = QHBoxLayout()
+        self.status_label = QLabel("就绪")
+        self.status_label.setStyleSheet("color: #666; font-size: 11px;")
+        status_layout.addWidget(self.status_label)
+        status_layout.addStretch()
+        layout.addLayout(status_layout)
+
+        self.setLayout(layout)
+
+        self.zoom_in_btn.clicked.connect(self.zoom_in)
+        self.zoom_out_btn.clicked.connect(self.zoom_out)
+        self.fit_btn.clicked.connect(self.fit_to_window)
+        self.original_btn.clicked.connect(self.original_size)
+        self.save_btn.clicked.connect(self.save_image)
+
+    def load_image(self, image_path: str):
+        """加载并显示图像"""
+        if not image_path or not Path(image_path).exists():
+            self.image_label.setText("图像不存在")
+            self.status_label.setText("图像加载失败")
+            return
+
+        self.current_image_path = image_path
+        self.scale_factor = 1.0
+
+        pixmap = QPixmap(image_path)
+        if pixmap.isNull():
+            self.image_label.setText("无法加载图像")
+            self.status_label.setText("图像格式不支持")
+            return
+
+        self.original_pixmap = pixmap
+        self.fit_to_window()
+
+        file_info = Path(image_path).stat()
+        size_mb = file_info.st_size / (1024 * 1024)
+        self.status_label.setText(f"{pixmap.width()}x{pixmap.height()} | {size_mb:.2f} MB | {Path(image_path).name} | 适应窗口")
+
+    def update_image_display(self):
+        """更新图像显示 - 使用防抖避免频繁重绘卡顿"""
+        self._update_timer.stop()
+        self._pending_scale = self.scale_factor
+        self._update_timer.start(50)
+
+    def _do_update_display(self):
+        """实际执行图像更新"""
+        if not hasattr(self, 'original_pixmap') or self.original_pixmap.isNull():
+            return
+
+        if self._pending_scale is None:
+            return
+
+        if self._pending_scale > 2.0 or self._pending_scale < 0.5:
+            transform = Qt.FastTransformation
+        else:
+            transform = Qt.SmoothTransformation
+
+        scaled_pixmap = self.original_pixmap.scaled(
+            int(self.original_pixmap.width() * self._pending_scale),
+            int(self.original_pixmap.height() * self._pending_scale),
+            Qt.KeepAspectRatio,
+            transform
+        )
+        self.image_label.setPixmap(scaled_pixmap)
+        self._pending_scale = None
+
+    def wheelEvent(self, event):
+        """鼠标滚轮缩放 - 实时响应"""
+        delta = event.angleDelta().y()
+
+        if delta > 0:
+            if self.scale_factor < 5.0:
+                self.scale_factor = min(self.scale_factor * 1.1, 5.0)
+                self.update_image_display()
+        else:
+            if self.scale_factor > 0.1:
+                self.scale_factor = max(self.scale_factor / 1.1, 0.1)
+                self.update_image_display()
+
+        event.accept()
+
+    def zoom_in(self):
+        """放大"""
+        if self.scale_factor < 5.0:
+            self.scale_factor = min(self.scale_factor * 1.25, 5.0)
+            self.update_image_display()
+
+    def zoom_out(self):
+        """缩小"""
+        if self.scale_factor > 0.1:
+            self.scale_factor = max(self.scale_factor / 1.25, 0.1)
+            self.update_image_display()
+
+    def fit_to_window(self):
+        """适应窗口"""
+        if not hasattr(self, 'original_pixmap') or self.original_pixmap.isNull():
+            return
+
+        view_size = self.scroll_area.viewport().size()
+        img_size = self.original_pixmap.size()
+
+        scale_w = view_size.width() / img_size.width()
+        scale_h = view_size.height() / img_size.height()
+
+        self._fit_scale = min(scale_w, scale_h)
+        self.scale_factor = self._fit_scale
+
+        self.update_image_display()
+        self.status_label.setText(f"适应窗口 | 缩放: {self.scale_factor:.1%}")
+
+    def original_size(self):
+        """原始大小"""
+        self.scale_factor = 1.0
+        self._fit_scale = None
+        self.update_image_display()
+        self.status_label.setText("原始大小 | 缩放: 100%")
+
+    def save_image(self):
+        """保存图像"""
+        if not self.current_image_path:
+            return
+
+        file_path, _ = QFileDialog.getSaveFileName(
+            self, "保存图像", Path(self.current_image_path).name,
+            "PNG图片 (*.png);;JPG图片 (*.jpg);;所有文件 (*.*)"
+        )
+
+        if file_path:
+            try:
+                import shutil
+                shutil.copy(self.current_image_path, file_path)
+            except Exception as e:
+                QMessageBox.critical(self, "错误", f"保存失败: {e}")

src/gui/core/config_manager.py

@@ -0,0 +1,161 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+配置管理器
+
+接管主窗口中所有配置读写逻辑：
+- new_config()           清空所有面板配置
+- load_config(file_path) 从 JSON 文件加载配置并回填面板
+- save_config(file_path) 将当前配置保存为 JSON 文件
+- get_current_config()   遍历 PanelFactory 收集配置（懒加载安全）
+
+懒加载兼容原则：
+- get_current_config() 仅遍历已加载面板，未加载面板返回空字典 {}
+- 绝不为了拿配置而强行唤醒/渲染所有 Panel
+- 如需全量配置（如保存），调用方应先执行 panel_factory.preload_all()
+"""
+
+import json
+import os
+from typing import Dict, Optional
+
+from PyQt5.QtCore import QObject
+from PyQt5.QtWidgets import QMessageBox, QFileDialog
+
+from src.gui.core.event_bus import global_event_bus
+
+
+class ConfigManager(QObject):
+    """配置管理器。
+
+    用法::
+
+        cfg_mgr = ConfigManager(panel_factory, parent=self)
+        cfg_mgr.new_config()          # 清空配置
+        cfg_mgr.load_config(path)     # 加载 JSON
+        cfg_mgr.save_config(path)     # 保存 JSON
+        config = cfg_mgr.get_current_config()  # 收集当前配置
+    """
+
+    def __init__(self, panel_factory, parent=None):
+        """
+        Args:
+            panel_factory: PanelFactory 实例
+            parent: 父 QObject（用于弹窗定位）
+        """
+        super().__init__(parent)
+        self._panel_factory = panel_factory
+
+    # ═══════════════════════════════════════════════════════════
+    # 公开 API
+    # ═══════════════════════════════════════════════════════════
+
+    def new_config(self):
+        """清空所有面板配置（需用户确认）。"""
+        reply = QMessageBox.question(
+            self.parent(), "新建配置", "是否清空当前配置？",
+            QMessageBox.Yes | QMessageBox.No
+        )
+        if reply != QMessageBox.Yes:
+            return
+
+        for panel in self._panel_factory.get_loaded_panels().values():
+            if hasattr(panel, 'clear_config'):
+                panel.clear_config()
+
+        global_event_bus.publish('LogMessage', {
+            'message': '已清空配置',
+            'level': 'info',
+        })
+
+    def load_config(self, file_path: str = None):
+        """从 JSON 文件加载配置并回填面板。
+
+        Args:
+            file_path: JSON 文件路径。若为 None，弹出文件选择对话框。
+        """
+        if file_path is None:
+            file_path, _ = QFileDialog.getOpenFileName(
+                self.parent(), "加载配置", "",
+                "JSON Files (*.json);;All Files (*.*)"
+            )
+            if not file_path:
+                return
+
+        try:
+            with open(file_path, 'r', encoding='utf-8') as f:
+                config = json.load(f)
+        except Exception as e:
+            QMessageBox.critical(
+                self.parent(), "加载失败",
+                f"无法读取配置文件:\n{file_path}\n\n错误: {e}"
+            )
+            return
+
+        # 回填已加载面板
+        loaded_count = 0
+        for step_id, panel in self._panel_factory.get_loaded_panels().items():
+            if step_id in config and hasattr(panel, 'set_config'):
+                try:
+                    panel.set_config(config[step_id])
+                    loaded_count += 1
+                except Exception:
+                    pass
+
+        global_event_bus.publish('LogMessage', {
+            'message': f'已加载配置: {file_path}（回填 {loaded_count} 个面板）',
+            'level': 'info',
+        })
+
+    def save_config(self, file_path: str = None):
+        """将当前配置保存为 JSON 文件。
+
+        注意：保存前会强制加载所有面板（preload_all），确保配置完整。
+
+        Args:
+            file_path: 目标 JSON 文件路径。若为 None，弹出保存对话框。
+        """
+        if file_path is None:
+            file_path, _ = QFileDialog.getSaveFileName(
+                self.parent(), "保存配置", "config.json",
+                "JSON Files (*.json);;All Files (*.*)"
+            )
+            if not file_path:
+                return
+
+        # 保存前强制加载所有面板，确保配置完整
+        self._panel_factory.preload_all()
+        config = self.get_current_config()
+
+        try:
+            with open(file_path, 'w', encoding='utf-8') as f:
+                json.dump(config, f, indent=2, ensure_ascii=False)
+        except Exception as e:
+            QMessageBox.critical(
+                self.parent(), "保存失败",
+                f"无法保存配置文件:\n{file_path}\n\n错误: {e}"
+            )
+            return
+
+        global_event_bus.publish('LogMessage', {
+            'message': f'已保存配置: {file_path}',
+            'level': 'info',
+        })
+
+    def get_current_config(self) -> Dict[str, dict]:
+        """收集当前所有步骤的配置。
+
+        懒加载安全：仅遍历已加载面板，未加载面板返回空字典 {}。
+        绝不为了拿配置而强行唤醒/渲染所有 Panel。
+
+        Returns:
+            {step_id: panel_config_dict}
+        """
+        config = {}
+        for step_id, panel in self._panel_factory.get_loaded_panels().items():
+            if hasattr(panel, 'get_config'):
+                try:
+                    config[step_id] = panel.get_config()
+                except Exception:
+                    config[step_id] = {}
+        return config

src/gui/core/dependency_subscriber.py

@@ -0,0 +1,64 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+依赖订阅混入模块
+
+提供 subscribe_panel_to_dependencies() 函数，让步骤面板根据
+PANEL_REGISTRY 中声明的 dependencies 自动向 global_event_bus
+订阅 OutputUpdated 事件。当上游步骤产出落地时，面板自动将路径
+填入对应的 FileSelectWidget，无需主窗口手工传导。
+"""
+
+from src.gui.core.event_bus import global_event_bus
+
+
+def subscribe_panel_to_dependencies(panel, step_id, dependencies):
+    """为面板订阅其依赖的上游步骤产出事件。
+
+    当 global_event_bus 发布 OutputUpdated 事件且 step_id/output_type
+    匹配时，自动将路径填入面板对应的 FileSelectWidget。
+
+    Args:
+        panel: 步骤面板实例（QWidget 子类）
+        step_id: 当前面板的 step_id（仅用于日志，非匹配键）
+        dependencies: dict, {input_field: (dep_step, output_type, panel_attr)}
+    """
+    if not dependencies:
+        return
+
+    for _input_field, (dep_step, output_type, panel_attr) in dependencies.items():
+        _make_subscription(panel, dep_step, output_type, panel_attr)
+
+
+def _make_subscription(panel, dep_step, output_type, panel_attr):
+    """为单个依赖项创建事件订阅。使用工厂函数避免闭包变量延迟绑定。"""
+
+    def callback(data):
+        if data.get('step_id') != dep_step:
+            return
+        if data.get('output_type') != output_type:
+            return
+
+        widget = getattr(panel, panel_attr, None)
+        if widget is None:
+            return
+
+        current = ''
+        if hasattr(widget, 'get_path'):
+            current = widget.get_path().strip()
+        elif hasattr(widget, 'text'):
+            current = widget.text().strip()
+
+        if current:
+            return
+
+        path = data.get('path', '')
+        if not path:
+            return
+
+        if hasattr(widget, 'set_path'):
+            widget.set_path(path)
+        elif hasattr(widget, 'setText'):
+            widget.setText(path)
+
+    global_event_bus.subscribe('OutputUpdated', callback)

src/gui/core/dialog_service.py

@@ -0,0 +1,67 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+对话框服务
+
+封装纯展示类弹窗，从主窗口中彻底剥离 UI 对话框逻辑。
+
+职责：
+- show_pipeline_status()  显示 Pipeline 模块加载状态
+- show_about()            显示"关于"对话框
+- show_ai_settings()      显示 AI 引擎配置对话框
+"""
+
+from PyQt5.QtCore import QObject
+from PyQt5.QtWidgets import QMessageBox
+
+
+class DialogService(QObject):
+    """对话框服务。
+
+    用法::
+
+        dlg_svc = DialogService(parent=self)
+        dlg_svc.show_about()
+        dlg_svc.show_pipeline_status()
+        dlg_svc.show_ai_settings()
+    """
+
+    def __init__(self, parent=None):
+        super().__init__(parent)
+
+    # ═══════════════════════════════════════════════════════════
+    # 公开 API
+    # ═══════════════════════════════════════════════════════════
+
+    def show_pipeline_status(self):
+        """显示 Pipeline 模块加载状态。"""
+        from src.gui.core.worker_thread import PIPELINE_AVAILABLE, PIPELINE_ERROR_INFO
+
+        if PIPELINE_AVAILABLE:
+            QMessageBox.information(
+                self.parent(), "Pipeline状态",
+                "Pipeline模块: 正常加载"
+            )
+        else:
+            detail = "\n".join(PIPELINE_ERROR_INFO)
+            QMessageBox.warning(
+                self.parent(), "Pipeline状态",
+                f"Pipeline模块: 加载失败\n\n{detail}"
+            )
+
+    def show_about(self):
+        """显示"关于"对话框。"""
+        QMessageBox.about(
+            self.parent(), "关于",
+            "MegaCube-Water Quality  V1.2\n\n"
+            "一个完整的水质参数反演工作流程工具\n\n"
+            "公司：北京依锐思遥感技术有限公司\n"
+            "地址：北京市海淀区清河安宁庄东路18号5号楼二层205\n"
+            "电话：010-51292601\n"
+            "邮箱：hanshanlong@iris-rs.cn"
+        )
+
+    def show_ai_settings(self):
+        """显示 AI 引擎配置对话框。"""
+        from src.gui.dialogs import AISettingsDialog
+        AISettingsDialog(self.parent()).exec_()

src/gui/core/event_bus.py

@@ -0,0 +1,34 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+轻量级事件总线
+
+支持 subscribe(event_name, callback) 和 publish(event_name, data)，
+用于步骤面板间的去中心化参数传导。
+"""
+
+from collections import defaultdict
+from typing import Any, Callable, Dict, List
+
+
+class EventBus:
+    """发布-订阅事件总线"""
+
+    def __init__(self):
+        self._subscribers: Dict[str, List[Callable]] = defaultdict(list)
+
+    def subscribe(self, event_name: str, callback: Callable[[dict], None]):
+        """订阅事件。callback 接收一个 dict 作为事件数据。"""
+        self._subscribers[event_name].append(callback)
+
+    def publish(self, event_name: str, data: Dict[str, Any]):
+        """发布事件，通知所有订阅者。"""
+        for callback in self._subscribers.get(event_name, []):
+            try:
+                callback(data)
+            except Exception:
+                pass
+
+
+# 全局单例
+global_event_bus = EventBus()

src/gui/core/log_manager.py

@@ -0,0 +1,187 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+日志与进度管理器
+
+将主窗口中的日志区（QTextEdit）、进度条（QProgressBar）和"清空日志"按钮
+的 UI 创建与控制逻辑完全封装。
+
+职责：
+- create_log_panel() → 返回组装好的 QWidget（日志 + 进度条）
+- 内部订阅 LogMessage / ProgressUpdate 事件，自动更新 UI
+- 主窗口无需再关心日志/进度的状态同步
+
+订阅的事件：
+  LogMessage      → {message, level}          写入日志区
+  ProgressUpdate  → {percentage, message}      更新进度条
+"""
+
+from datetime import datetime
+
+from PyQt5.QtCore import QObject
+from PyQt5.QtGui import QTextCursor
+from PyQt5.QtWidgets import (
+    QWidget, QVBoxLayout, QHBoxLayout,
+    QGroupBox, QTextEdit, QProgressBar, QPushButton,
+)
+
+from src.gui.core.event_bus import global_event_bus
+
+
+class LogManager(QObject):
+    """日志与进度管理器。
+
+    用法::
+
+        log_mgr = LogManager(parent=self)
+        log_panel = log_mgr.create_log_panel()
+        layout.addWidget(log_panel)
+        # 之后所有日志/进度更新由 EventBus 自动驱动，无需手动操作
+    """
+
+    def __init__(self, parent=None):
+        super().__init__(parent)
+        self._log_text: QTextEdit = None
+        self._progress_bar: QProgressBar = None
+
+        # 订阅事件
+        global_event_bus.subscribe('LogMessage', self._on_log_message)
+        global_event_bus.subscribe('ProgressUpdate', self._on_progress_update)
+
+    # ═══════════════════════════════════════════════════════════
+    # 公开 API
+    # ═══════════════════════════════════════════════════════════
+
+    def create_log_panel(self) -> QWidget:
+        """创建并返回日志+进度面板的组装 Widget。
+
+        Returns:
+            QWidget: 包含日志区（QGroupBox）和进度条（QGroupBox）的垂直布局容器
+        """
+        from src.gui.styles import ModernStylesheet
+
+        container = QWidget()
+        layout = QVBoxLayout()
+        layout.setContentsMargins(0, 0, 0, 0)
+        layout.setSpacing(10)
+
+        # ── 日志区 ──
+        log_group = QGroupBox("执行日志")
+        log_group.setStyleSheet(f"""
+            QGroupBox {{
+                background-color: {ModernStylesheet.COLORS['panel_bg']};
+                border: 1px solid {ModernStylesheet.COLORS['border_light']};
+                border-radius: 5px; margin-top: 8px; padding-top: 15px;
+                padding-left: 9px; padding-right: 9px; padding-bottom: 9px;
+            }}
+            QGroupBox::title {{
+                subcontrol-origin: margin; subcontrol-position: top left;
+                padding: 0 5px; font-weight: bold;
+                color: {ModernStylesheet.COLORS['text_primary']};
+            }}
+        """)
+        log_layout = QVBoxLayout()
+        log_layout.setContentsMargins(5, 5, 5, 5)
+
+        self._log_text = QTextEdit()
+        self._log_text.setReadOnly(True)
+        self._log_text.setMaximumHeight(200)
+        self._log_text.setStyleSheet(f"""
+            QTextEdit {{
+                background-color: {ModernStylesheet.COLORS['panel_bg']};
+                color: {ModernStylesheet.COLORS['text_primary']};
+                border: 1px solid {ModernStylesheet.COLORS['border']};
+                border-radius: 4px; padding: 5px;
+                font-family: 'Courier New', monospace; font-size: 10px;
+            }}
+        """)
+        log_layout.addWidget(self._log_text)
+
+        clear_btn = QPushButton("清空日志")
+        clear_btn.setMaximumWidth(100)
+        clear_btn.setStyleSheet(ModernStylesheet.get_button_stylesheet('normal'))
+        clear_btn.clicked.connect(self.clear_log)
+
+        btn_row = QHBoxLayout()
+        btn_row.addWidget(clear_btn)
+        btn_row.addStretch()
+        log_layout.addLayout(btn_row)
+
+        log_group.setLayout(log_layout)
+        layout.addWidget(log_group, 1)
+
+        # ── 进度条 ──
+        progress_group = QGroupBox("执行进度")
+        progress_group.setStyleSheet(f"""
+            QGroupBox {{
+                background-color: {ModernStylesheet.COLORS['panel_bg']};
+                border: 1px solid {ModernStylesheet.COLORS['border_light']};
+                border-radius: 5px; margin-top: 8px; padding-top: 10px;
+                padding-left: 9px; padding-right: 9px; padding-bottom: 9px;
+            }}
+            QGroupBox::title {{
+                subcontrol-origin: margin; subcontrol-position: top left;
+                padding: 0 5px; font-weight: bold;
+                color: {ModernStylesheet.COLORS['text_primary']};
+            }}
+        """)
+        progress_layout = QVBoxLayout()
+        progress_layout.setContentsMargins(5, 5, 5, 5)
+
+        self._progress_bar = QProgressBar()
+        self._progress_bar.setValue(0)
+        self._progress_bar.setStyleSheet(f"""
+            QProgressBar {{
+                background-color: {ModernStylesheet.COLORS['panel_bg']};
+                border: 1px solid {ModernStylesheet.COLORS['border']};
+                border-radius: 4px; padding: 2px; text-align: center; height: 20px;
+            }}
+            QProgressBar::chunk {{
+                background-color: {ModernStylesheet.COLORS['success']}; border-radius: 3px;
+            }}
+        """)
+        progress_layout.addWidget(self._progress_bar)
+        progress_group.setLayout(progress_layout)
+        layout.addWidget(progress_group, 0)
+
+        container.setLayout(layout)
+        return container
+
+    def clear_log(self):
+        """清空日志区。"""
+        if self._log_text is not None:
+            self._log_text.clear()
+
+    @property
+    def progress_bar(self) -> QProgressBar:
+        return self._progress_bar
+
+    @property
+    def log_text(self) -> QTextEdit:
+        return self._log_text
+
+    # ═══════════════════════════════════════════════════════════
+    # EventBus 订阅回调
+    # ═══════════════════════════════════════════════════════════
+
+    def _on_log_message(self, data: dict):
+        """LogMessage 事件回调：写入日志区。"""
+        if self._log_text is None:
+            return
+        message = data.get('message', '')
+        level = data.get('level', 'info')
+        timestamp = datetime.now().strftime('%Y-%m-%d %H:%M:%S')
+        color_map = {'error': 'red', 'warning': 'orange'}
+        color = color_map.get(level, 'black')
+        formatted = f'<span style="color: {color};">[{timestamp}] {message}</span>'
+        self._log_text.append(formatted)
+        cursor = self._log_text.textCursor()
+        cursor.movePosition(QTextCursor.End)
+        self._log_text.setTextCursor(cursor)
+
+    def _on_progress_update(self, data: dict):
+        """ProgressUpdate 事件回调：更新进度条。"""
+        if self._progress_bar is None:
+            return
+        percentage = data.get('percentage', 0)
+        self._progress_bar.setValue(percentage)

src/gui/core/panel_factory.py

@@ -0,0 +1,294 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+面板注册与装载工厂
+
+按需懒加载步骤面板，替代 create_content_area 中一次性全量 new 14 个面板的做法。
+主窗口只需持有 PanelFactory 实例，通过 factory.create_tab_widget() 获取
+已挂载占位页的 QTabWidget，面板在用户首次切换到对应 Tab 时才实例化。
+
+特性：
+- 懒加载：仅在 tab 首次激活时创建面板实例
+- 邻接预加载：切换 tab 时自动预加载左右邻居（可配置预加载窗口大小）
+- 注册表驱动：完全依赖 PANEL_REGISTRY，零硬编码
+- 事件总线自动接线：面板创建后自动调用 subscribe_panel_to_dependencies
+- 占位页：未加载的 tab 显示空白 QWidget，加载后原地替换为 QScrollArea(panel)
+"""
+
+from PyQt5.QtWidgets import QWidget, QTabWidget, QScrollArea
+from PyQt5.QtCore import Qt
+
+from src.gui.core.panel_registry import PANEL_REGISTRY
+from src.gui.core.dependency_subscriber import subscribe_panel_to_dependencies
+from src.gui.core.event_bus import global_event_bus
+
+
+class PanelFactory:
+    """面板注册与装载工厂。
+
+    用法::
+
+        factory = PanelFactory(registry=PANEL_REGISTRY, main_window=self)
+        tab_widget = factory.create_tab_widget(icons_dir="data/icons")
+        # tab_widget 已包含所有占位页，可直接加入主窗口布局
+        # 后续通过 factory.get_panel(step_id) 按需获取面板实例
+    """
+
+    def __init__(self, registry, main_window, preload_window=1):
+        """
+        Args:
+            registry: PANEL_REGISTRY 列表
+            main_window: WaterQualityGUI 实例（用于注入 main_window 依赖）
+            preload_window: 邻接预加载窗口大小。0=仅加载当前 tab；
+                           1=当前+左右各1个；-1=全量预加载（退化为旧行为）
+        """
+        self._registry = registry
+        self._main_window = main_window
+        self._preload_window = preload_window
+
+        # step_id → panel 实例（仅已加载的）
+        self._panels = {}
+        # tab_index → 是否已加载
+        self._loaded = set()
+        # tab_index → placeholder QWidget（加载后被替换）
+        self._placeholders = {}
+        # 对外的 QTabWidget 引用
+        self._tab_widget = None
+
+    # ── 公开 API ──────────────────────────────────────────────
+
+    def create_tab_widget(self, icons_dir="data/icons"):
+        """创建并返回已填充占位页的 QTabWidget。
+
+        每个 tab 初始为空白 QWidget 占位，面板在首次激活时懒加载。
+        同时连接 currentChanged 信号驱动懒加载 + 邻接预加载。
+
+        Args:
+            icons_dir: 图标目录名（相对于项目根），用于 get_resource_path
+
+        Returns:
+            QTabWidget: 已添加所有占位 tab 的标签页控件
+        """
+        from src.gui.water_quality_gui import get_resource_path
+        from PyQt5.QtGui import QIcon
+
+        self._tab_widget = QTabWidget()
+        self._tab_widget.setTabPosition(QTabWidget.North)
+        self._tab_widget.setTabsClosable(False)
+
+        for idx, entry in enumerate(self._registry):
+            step_id = entry['step_id']
+            title = entry['title']
+            icon_name = entry['icon']
+
+            # 创建占位页
+            placeholder = QWidget()
+            self._placeholders[idx] = placeholder
+
+            icon_path = get_resource_path(f"{icons_dir}/{icon_name}")
+            self._tab_widget.addTab(placeholder, QIcon(icon_path), title)
+
+        # 连接切换信号 → 懒加载
+        self._tab_widget.currentChanged.connect(self._on_tab_changed)
+
+        # 立即预加载首个 tab
+        if self._registry:
+            self._ensure_loaded(0)
+
+        return self._tab_widget
+
+    def get_panel(self, step_id):
+        """获取面板实例（若未加载则触发懒加载）。
+
+        Args:
+            step_id: 步骤 ID，如 'step1'、'step5_clean'
+
+        Returns:
+            QWidget 或 None: 面板实例，未找到则返回 None
+        """
+        tab_index = self._step_id_to_tab_index(step_id)
+        if tab_index < 0:
+            return None
+        self._ensure_loaded(tab_index)
+        return self._panels.get(step_id)
+
+    def get_loaded_panels(self):
+        """返回所有已加载的面板字典 {step_id: panel}。"""
+        return dict(self._panels)
+
+    def preload_all(self):
+        """强制加载所有面板（用于配置保存等需要遍历全部面板的场景）。"""
+        for idx in range(len(self._registry)):
+            self._ensure_loaded(idx)
+
+    def get_tab_widget(self):
+        """返回内部 QTabWidget 引用。"""
+        return self._tab_widget
+
+    # ── 内部方法 ──────────────────────────────────────────────
+
+    def _on_tab_changed(self, index):
+        """Tab 切换时：加载当前 tab + 邻接预加载。"""
+        if index < 0:
+            return
+        self._ensure_loaded(index)
+        self._preload_neighbors(index)
+
+    def _ensure_loaded(self, tab_index):
+        """确保指定 tab 已加载；若未加载则实例化面板并替换占位页。"""
+        if tab_index in self._loaded:
+            return
+        if tab_index < 0 or tab_index >= len(self._registry):
+            return
+
+        entry = self._registry[tab_index]
+        step_id = entry['step_id']
+        cls = entry['class_ref']
+        title = entry['title']
+        kwargs = entry.get('constructor_kwargs')
+        deps = entry.get('dependencies')
+
+        # 解析构造参数
+        resolved_kwargs = {}
+        if kwargs:
+            for k in kwargs:
+                if k == 'main_window':
+                    resolved_kwargs[k] = self._main_window
+
+        # 实例化面板
+        panel = cls(**resolved_kwargs)
+
+        # 包裹到 QScrollArea
+        scroll = QScrollArea()
+        scroll.setWidget(panel)
+        scroll.setWidgetResizable(True)
+
+        # 替换占位页（blockSignals 阻断 removeTab/insertTab/setCurrentIndex 触发的
+        # currentChanged 信号风暴，防止 _on_tab_changed → _ensure_loaded 无限递归）
+        placeholder = self._placeholders.get(tab_index)
+        if placeholder is not None and self._tab_widget is not None:
+            tab_title = self._tab_widget.tabText(tab_index)
+            tab_icon = self._tab_widget.tabIcon(tab_index)
+            self._tab_widget.blockSignals(True)
+            try:
+                self._tab_widget.removeTab(tab_index)
+                self._tab_widget.insertTab(tab_index, scroll, tab_icon, tab_title)
+                self._tab_widget.setCurrentIndex(tab_index)
+            finally:
+                self._tab_widget.blockSignals(False)
+
+        # 注册
+        self._panels[step_id] = panel
+        self._loaded.add(tab_index)
+
+        # 事件总线自动接线
+        if deps:
+            subscribe_panel_to_dependencies(panel, step_id, deps)
+
+        # ★ Catch-up：向刚苏醒的懒加载面板回放已累积的状态
+        #   （面板在 OutputUpdated 事件广播之后才实例化，错过了事件，
+        #     必须主动回放 step_outputs + 全局输入，否则输入框全空）
+        self._replay_state_to_panel(panel)
+
+    # ── Catch-up 状态追溯 ────────────────────────────────────
+
+    def _replay_state_to_panel(self, panel):
+        """向刚实例化的懒加载面板回放已累积的状态。
+
+        三步回放：
+        1. update_from_config —— 生成默认输出路径 + 跨面板参数读取
+        2. 回放 WorkspaceManager.step_outputs —— 已运行步骤的产出文件路径
+        3. 实时扫描已加载面板 —— 读取被依赖的属性值（如 Step1 的 img_file）
+           发布为 OutputUpdated，触发 dependency_subscriber 回填输入框
+        """
+        # 1. update_from_config：生成默认输出路径
+        if hasattr(panel, 'update_from_config'):
+            try:
+                work_dir = self._get_current_work_dir()
+                panel.update_from_config(work_dir=work_dir, pipeline=None)
+            except Exception:
+                pass
+
+        # 2. 回放 WorkspaceManager 中已累积的 step_outputs
+        ws_manager = self._get_workspace_manager()
+        if ws_manager:
+            for src_step_id, outputs in ws_manager.step_outputs.items():
+                for output_type, path in outputs.items():
+                    if not path:
+                        continue
+                    global_event_bus.publish('OutputUpdated', {
+                        'step_id': src_step_id,
+                        'output_type': output_type,
+                        'path': path,
+                    })
+
+        # 3. 实时扫描已加载面板中被依赖的属性（覆盖全局输入如 reference_img）
+        self._replay_live_panel_inputs()
+
+    def _replay_live_panel_inputs(self):
+        """遍历 PANEL_REGISTRY 依赖声明，从已加载面板实时读取属性值。
+
+        若源面板已实例化，读取其 widget 的当前值并发布为 OutputUpdated，
+        确保懒加载面板能收到全局输入（如 Step1.img_file → reference_img）。
+        """
+        for entry in self._registry:
+            deps = entry.get('dependencies')
+            if not deps:
+                continue
+            for _input_field, (dep_step, output_type, panel_attr) in deps.items():
+                src_panel = self._panels.get(dep_step)
+                if src_panel is None:
+                    continue
+                widget = getattr(src_panel, panel_attr, None)
+                if widget is None:
+                    continue
+                path = ''
+                if hasattr(widget, 'get_path'):
+                    path = widget.get_path().strip()
+                elif hasattr(widget, 'text'):
+                    path = widget.text().strip()
+                if not path:
+                    continue
+                global_event_bus.publish('OutputUpdated', {
+                    'step_id': dep_step,
+                    'output_type': output_type,
+                    'path': path,
+                })
+
+    def _get_current_work_dir(self):
+        """从 WorkspaceInitializer 获取当前工作目录。"""
+        try:
+            return self._main_window._workspace_initializer.work_dir
+        except Exception:
+            return None
+
+    def _get_workspace_manager(self):
+        """从 WorkspaceInitializer 获取 WorkspaceManager 实例。"""
+        try:
+            return self._main_window._workspace_initializer.workspace_manager
+        except Exception:
+            return None
+
+    def _preload_neighbors(self, index):
+        """预加载当前 tab 的邻居（根据 preload_window 配置）。"""
+        if self._preload_window < 0:
+            # 全量预加载
+            for i in range(len(self._registry)):
+                self._ensure_loaded(i)
+            return
+
+        if self._preload_window == 0:
+            return
+
+        start = max(0, index - self._preload_window)
+        end = min(len(self._registry), index + self._preload_window + 1)
+        for i in range(start, end):
+            if i != index:
+                self._ensure_loaded(i)
+
+    def _step_id_to_tab_index(self, step_id):
+        """step_id → tab_index 映射。"""
+        for i, entry in enumerate(self._registry):
+            if entry['step_id'] == step_id:
+                return i
+        return -1

src/gui/core/panel_registry.py

@@ -0,0 +1,270 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+面板注册中心
+
+集中定义所有步骤面板的结构化配置，包括：
+- 步骤 ID / 类引用 / 标题 / 图标 / 阶段 / 导航显示名
+- 步骤间依赖关系（输入字段 → 上游步骤/输出类型/面板属性）
+- 构造参数（如 Step13ReportPanel 需要 main_window）
+
+WaterQualityGUI 通过遍历 PANEL_REGISTRY 动态生成导航树、Tab 页、
+依赖传递和配置读写，彻底消除硬编码。
+"""
+
+from src.gui.panels.step1_panel import Step1Panel
+from src.gui.panels.step2_panel import Step2Panel
+from src.gui.panels.step3_panel import Step3Panel
+from src.gui.panels.step4_sampling_panel import Step4SamplingPanel
+from src.gui.panels.step5_clean_panel import Step5CleanPanel
+from src.gui.panels.step6_feature_panel import Step6FeaturePanel
+from src.new.views.step7_view import Step7View
+from src.gui.panels.step8_ml_train_panel import Step8MlTrainPanel
+from src.gui.panels.step9_ml_predict_panel import Step9MlPredictPanel
+from src.gui.panels.step10_watercolor_panel import Step10WatercolorPanel
+from src.gui.panels.step11_map_panel import Step11MapPanel
+from src.gui.panels.step12_viz_panel import Step12VizPanel
+from src.gui.panels.step13_report_panel import Step13ReportPanel
+
+
+PANEL_REGISTRY = [
+    # ═══════════════════════════════════════════════════════════════
+    # 阶段一：影像预处理
+    # ═══════════════════════════════════════════════════════════════
+    {
+        'step_id': 'step1',
+        'class_ref': Step1Panel,
+        'title': '水域掩膜',
+        'icon': '1.png',
+        'stage': '阶段一：影像预处理',
+        'display_name': '1. 水域掩膜生成',
+        'dependencies': None,
+        'constructor_kwargs': None,
+    },
+    {
+        'step_id': 'step2',
+        'class_ref': Step2Panel,
+        'title': '耀斑检测',
+        'icon': '2.png',
+        'stage': '阶段一：影像预处理',
+        'display_name': '2. 耀斑区域识别',
+        'dependencies': {
+            'img_path': ('step1', 'reference_img', 'img_file'),
+            'water_mask_path': ('step1', 'water_mask', 'water_mask_file'),
+        },
+        'constructor_kwargs': None,
+    },
+    {
+        'step_id': 'step3',
+        'class_ref': Step3Panel,
+        'title': '耀斑去除',
+        'icon': '3.png',
+        'stage': '阶段一：影像预处理',
+        'display_name': '3. 耀斑去除与修复',
+        'dependencies': {
+            'img_path': ('step1', 'reference_img', 'img_file'),
+            'water_mask': ('step1', 'water_mask', 'water_mask_file'),
+        },
+        'constructor_kwargs': None,
+    },
+
+    # ═══════════════════════════════════════════════════════════════
+    # 阶段二：样本数据准备
+    # ═══════════════════════════════════════════════════════════════
+    {
+        'step_id': 'step4_sampling',
+        'class_ref': Step4SamplingPanel,
+        'title': '采样点布设',
+        'icon': '4.png',
+        'stage': '阶段二：样本数据准备',
+        'display_name': '4. 采样点布设',
+        'dependencies': {
+            'deglint_img_path': ('step3', 'deglint_image', 'deglint_img_file'),
+            'water_mask_path': ('step1', 'water_mask', 'water_mask_file'),
+        },
+        'constructor_kwargs': None,
+    },
+    {
+        'step_id': 'step5_clean',
+        'class_ref': Step5CleanPanel,
+        'title': '数据清洗',
+        'icon': '5.png',
+        'stage': '阶段二：样本数据准备',
+        'display_name': '5. 数据清洗',
+        # 业务要求保持输入源独立，不自动抓取 step4_sampling 的输出
+        'dependencies': None,
+        'constructor_kwargs': None,
+    },
+    {
+        'step_id': 'step6_feature',
+        'class_ref': Step6FeaturePanel,
+        'title': '光谱特征',
+        'icon': '6.png',
+        'stage': '阶段二：样本数据准备',
+        'display_name': '6. 光谱特征提取',
+        'dependencies': {
+            'deglint_img_path': ('step3', 'deglint_image', 'deglint_img_file'),
+            'csv_path': ('step5_clean', 'processed_data', 'csv_file'),
+            'boundary_mask_path': ('step1', 'water_mask', 'water_mask_file'),
+            'glint_mask_path': ('step2', 'glint_mask', 'glint_mask_file'),
+        },
+        'constructor_kwargs': None,
+    },
+    {
+        'step_id': 'step7_index',
+        'class_ref': Step7View,
+        'title': '水质光谱指数计算',
+        'icon': '7.png',
+        'stage': '阶段二：样本数据准备',
+        'display_name': '7. 水质指数计算',
+        'dependencies': {
+            'training_csv_path': ('step6_feature', 'training_spectra', 'training_data_widget'),
+        },
+        'constructor_kwargs': None,
+    },
+
+    # ═══════════════════════════════════════════════════════════════
+    # 阶段三：模型构建与训练
+    # ═══════════════════════════════════════════════════════════════
+    {
+        'step_id': 'step8_ml_train',
+        'class_ref': Step8MlTrainPanel,
+        'title': '机器学习建模',
+        'icon': '8.png',
+        'stage': '阶段三：模型构建与训练',
+        'display_name': '8. 机器学习建模',
+        'dependencies': {
+            'training_csv_file': ('step7_index', 'training_spectra_indices', 'training_csv_file'),
+        },
+        'constructor_kwargs': None,
+    },
+
+    # ═══════════════════════════════════════════════════════════════
+    # 阶段四：预测与成果输出
+    # ═══════════════════════════════════════════════════════════════
+    {
+        'step_id': 'step9_ml_predict',
+        'class_ref': Step9MlPredictPanel,
+        'title': '机器学习预测',
+        'icon': '10.png',
+        'stage': '阶段四：预测与成果输出',
+        'display_name': '9. 机器学习预测',
+        'dependencies': {
+            'models_dir': ('step8_ml_train', 'Supervised_Model_Training', 'models_dir_file'),
+        },
+        'constructor_kwargs': None,
+    },
+    {
+        'step_id': 'step10_watercolor',
+        'class_ref': Step10WatercolorPanel,
+        'title': '水色指数反演',
+        'icon': '10.png',
+        'stage': '阶段四：预测与成果输出',
+        'display_name': '10. 水色指数反演',
+        'dependencies': {
+            'bsq_file': ('step3', 'deglint_image', 'bsq_file'),
+        },
+        'constructor_kwargs': None,
+    },
+    {
+        'step_id': 'step11_map',
+        'class_ref': Step11MapPanel,
+        'title': '专题图生成',
+        'icon': '10.png',
+        'stage': '阶段四：预测与成果输出',
+        'display_name': '11. 专题图生成',
+        'dependencies': {
+            'prediction_csv_dir_edit': ('step9_ml_predict', '9_ML_Prediction', 'prediction_csv_dir_edit'),
+            'geotiff_dir_edit': ('step10_watercolor', 'WaterIndex_Images', 'geotiff_dir_edit'),
+        },
+        'constructor_kwargs': None,
+    },
+    {
+        'step_id': 'step12_viz',
+        'class_ref': Step12VizPanel,
+        'title': '可视化',
+        'icon': '9.png',
+        'stage': '阶段四：预测与成果输出',
+        'display_name': '12. 可视化展示',
+        'dependencies': None,
+        'constructor_kwargs': None,
+    },
+    {
+        'step_id': 'step13_report',
+        'class_ref': Step13ReportPanel,
+        'title': '报告生成',
+        'icon': '10.png',
+        'stage': '阶段四：预测与成果输出',
+        'display_name': '13. 分析报告生成',
+        'dependencies': None,
+        'constructor_kwargs': {'main_window'},  # 需要注入 main_window=self
+    },
+]
+
+
+def build_step_dependencies():
+    """从 PANEL_REGISTRY 构建 step_dependencies 字典。
+
+    Returns:
+        dict: {step_id: {input_field: (dep_step, output_type, panel_attr)}}
+    """
+    deps = {}
+    for entry in PANEL_REGISTRY:
+        if entry['dependencies']:
+            deps[entry['step_id']] = entry['dependencies']
+    return deps
+
+
+def build_stage_groups():
+    """从 PANEL_REGISTRY 构建阶段分组字典。
+
+    Returns:
+        dict: {stage_name: [(step_id, display_name), ...]}
+    """
+    groups = {}
+    for entry in PANEL_REGISTRY:
+        stage = entry['stage']
+        if stage not in groups:
+            groups[stage] = []
+        groups[stage].append((entry['step_id'], entry['display_name']))
+    return groups
+
+
+def get_tab_index(step_id):
+    """根据 step_id 获取其在 PANEL_REGISTRY 中的索引（即 Tab 索引）。"""
+    for i, entry in enumerate(PANEL_REGISTRY):
+        if entry['step_id'] == step_id:
+            return i
+    return -1
+
+
+def get_step_id_by_tab_index(tab_index):
+    """根据 Tab 索引获取 step_id。"""
+    if 0 <= tab_index < len(PANEL_REGISTRY):
+        return PANEL_REGISTRY[tab_index]['step_id']
+    return None
+
+
+def get_entry(step_id):
+    """根据 step_id 获取注册表条目。"""
+    for entry in PANEL_REGISTRY:
+        if entry['step_id'] == step_id:
+            return entry
+    return None
+
+
+def build_output_types():
+    """从 PANEL_REGISTRY 的 dependencies 反向推导每个步骤产出的 output_type 列表。
+
+    Returns:
+        dict: {step_id: [output_type, ...]}
+    """
+    output_types = {}
+    for entry in PANEL_REGISTRY:
+        if entry['dependencies']:
+            for _input_field, (dep_step, output_type, _panel_attr) in entry['dependencies'].items():
+                if dep_step not in output_types:
+                    output_types[dep_step] = []
+                if output_type not in output_types[dep_step]:
+                    output_types[dep_step].append(output_type)
+    return output_types

src/gui/core/pipeline_executor.py

@@ -0,0 +1,622 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+Pipeline 执行器
+
+接管 WaterQualityGUI 中所有 Pipeline 执行相关逻辑：
+- run_full_pipeline()  完整流程执行
+- run_single_step()    单步执行
+- stop_pipeline()      停止执行
+- _precheck_step3_bands()  step3 波段越界预检
+
+关键设计原则：
+- 所有状态变化通过 global_event_bus 发布事件，绝不直接操作 UI 控件
+- WorkerThread 的 Qt 信号连接到内部槽函数，槽函数仅做 EventBus 转发
+- 预检对话框（PreflightDialog / PipelineModeDialog / BandConfirmDialog）
+  仍为模态弹窗（用户交互必需），但结果通过 EventBus 发布
+
+发布的事件：
+  PipelineStarted    → {}                        主窗口订阅：禁用运行按钮
+  PipelineFinished   → {success, message}         主窗口订阅：恢复按钮 + 弹窗
+  PipelineStopped    → {}                        主窗口订阅：恢复按钮
+  StepCompleted      → {step_name, success, message}  WorkspaceInitializer 订阅：扫描产物
+  LogMessage         → {message, level}           LogManager 订阅：写入日志区
+  ProgressUpdate     → {percentage, message}      LogManager 订阅：更新进度条
+"""
+
+import os
+import copy
+import traceback
+from pathlib import Path
+from typing import Dict, List, Optional
+
+from PyQt5.QtCore import QObject, Qt
+from PyQt5.QtWidgets import QMessageBox, QDialog
+
+from src.gui.core.event_bus import global_event_bus
+from src.gui.core.worker_thread import (
+    WorkerThread,
+    PIPELINE_AVAILABLE,
+)
+from src.gui.core.preflight_dialog import PreflightDialog
+from src.gui.core.pipeline_mode_dialog import PipelineModeDialog
+from src.gui.dialogs import BandConfirmDialog
+from src.core.pipeline.runner import PipelineHalt
+
+class PipelineExecutor(QObject):
+    """Pipeline 执行器 —— 纯逻辑层，零 UI 直接操作。"""
+
+    def __init__(self, panel_factory, workspace_initializer, parent=None):
+        """
+        Args:
+            panel_factory: PanelFactory 实例（用于获取面板和配置）
+            workspace_initializer: WorkspaceInitializer 实例（用于获取 work_dir）
+            parent: 父 QObject（通常为 WaterQualityGUI）
+        """
+        super().__init__(parent)
+        self._panel_factory = panel_factory
+        self._workspace_initializer = workspace_initializer
+        self._worker: Optional[WorkerThread] = None
+
+        # 订阅面板发出的单步执行请求（解耦面板与执行器）
+        global_event_bus.subscribe('RequestRunSingleStep', self._on_request_run_single_step)
+
+    # ═══════════════════════════════════════════════════════════
+    # 公开 API
+    # ═══════════════════════════════════════════════════════════
+
+    @property
+    def worker(self):
+        return self._worker
+
+    @property
+    def is_running(self) -> bool:
+        return self._worker is not None and self._worker.isRunning()
+
+    def run_full_pipeline(self):
+        """运行完整流程。
+
+        流程：
+        1. 检查 PIPELINE_AVAILABLE
+        2. 获取 work_dir（从 WorkspaceInitializer）
+        3. 扫描工作目录 + 自动回填
+        4. step3 波段越界预检
+        5. 全流程模式选择弹窗
+        6. 获取配置 + 模式裁剪
+        7. 一次性全预检 + 用户交互
+        8. 确认执行 → 创建 WorkerThread → 启动
+
+        关键防静默失败设计：
+        - 每一个 return 前必须通过 EventBus 发布 LogMessage
+        - 整个方法体包裹在 try/except 中，防止 PyQt5 槽函数静默吞异常
+        """
+        print("==== [探针] run_full_pipeline 方法体已进入 ====", flush=True)
+        try:
+            self._run_full_pipeline_impl()
+        except Exception as e:
+            err_detail = traceback.format_exc()
+            global_event_bus.publish('LogMessage', {
+                'message': f'[致命错误] run_full_pipeline 异常: {e}',
+                'level': 'error',
+            })
+            global_event_bus.publish('LogMessage', {
+                'message': f'详细追踪:\n{err_detail}',
+                'level': 'error',
+            })
+            QMessageBox.critical(
+                self.parent(), "运行失败",
+                f"启动流程时发生未预期的错误:\n\n{e}\n\n详细信息已输出到日志区。"
+            )
+
+    def _run_full_pipeline_impl(self):
+        """run_full_pipeline 的实现体，由外层 try/except 保护。"""
+        # ★ 终端即时反馈：确保即使 EventBus/日志区未就绪也能看到
+        print("\n[PipelineExecutor] 收到「运行完整流程」指令，开始执行...")
+
+        if not PIPELINE_AVAILABLE:
+            global_event_bus.publish('LogMessage', {
+                'message': '无法导入 Pipeline 模块，请检查项目文件结构！',
+                'level': 'error',
+            })
+            QMessageBox.critical(
+                self.parent(), "错误",
+                "无法导入 Pipeline 模块，请检查 src/core/handlers/ 目录是否完整！"
+            )
+            return
+
+        # ── 1) 获取 work_dir ──
+        work_dir = self._workspace_initializer.work_dir
+        if not work_dir:
+            global_event_bus.publish('LogMessage', {
+                'message': '⚠ 未选择工作目录，流程中止。请先通过「工具 → 设置工作目录」选择工作目录。',
+                'level': 'warning',
+            })
+            QMessageBox.warning(self.parent(), "警告", "未选择工作目录，请先设置工作目录。")
+            return
+
+        work_path = Path(work_dir)
+        global_event_bus.publish('LogMessage', {
+            'message': f'[运行] 工作目录: {work_dir}',
+            'level': 'info',
+        })
+
+        # ── 2) 运行前扫描 + 自动回填 ──
+        global_event_bus.publish('LogMessage', {
+            'message': '正在进行运行前环境预检与自动扫描...',
+            'level': 'info',
+        })
+        self._workspace_initializer.auto_populate_all()
+        global_event_bus.publish('LogMessage', {
+            'message': '✓ 预检完成：已扫描工作目录并自动回填已落盘的产物',
+            'level': 'info',
+        })
+
+        # ── 3) step3 波段越界预检 ──
+        if not self._precheck_step3_bands():
+            global_event_bus.publish('LogMessage', {
+                'message': '⚠ 流程中止：step3 波段越界预检未通过（用户取消或波段配置无效）',
+                'level': 'warning',
+            })
+            return
+
+        # ── 4) 全流程模式选择弹窗 ──
+        mode_dlg = PipelineModeDialog(main_window=self.parent(), parent=self.parent())
+        if mode_dlg.exec() != QDialog.Accepted:
+            global_event_bus.publish('LogMessage', {
+                'message': '⚠ 流程中止：用户取消了模式选择对话框',
+                'level': 'warning',
+            })
+            return
+        selected_mode = mode_dlg.selected_mode
+        global_event_bus.publish('LogMessage', {
+            'message': (
+                f"[模式选择] 选定模式: "
+                f"{'训练新模型' if selected_mode == 'training' else '使用已有模型直接预测'}"
+            ),
+            'level': 'info',
+        })
+
+        # ── 5) 获取配置（★ 先预加载所有面板，确保配置完整） ──
+        global_event_bus.publish('LogMessage', {
+            'message': '[运行] 正在收集所有步骤面板的配置...',
+            'level': 'info',
+        })
+        self._panel_factory.preload_all()
+        config = self._get_current_config()
+        global_event_bus.publish('LogMessage', {
+            'message': f'[运行] 已收集 {len(config)} 个步骤的配置: {list(config.keys())}',
+            'level': 'info',
+        })
+
+        # ── 6) 模式裁剪 ──
+        if selected_mode == "prediction_only":
+            from src.core.workspace_manager import WorkspaceManager
+            config = WorkspaceManager.prune_config_for_prediction_mode(config)
+            global_event_bus.publish('LogMessage', {
+                'message': '[模式选择] 已裁剪训练相关步骤（step4/5/7/8），进入仅预测模式',
+                'level': 'info',
+            })
+
+        # ── 7) 一次性全预检 + 用户交互式决策 ──
+        missing_items = PreflightDialog.build_missing_items(config)
+        skip_list: List[str] = []
+
+        if missing_items:
+            global_event_bus.publish('LogMessage', {
+                'message': f'[预检] 发现 {len(missing_items)} 个缺失项，弹出预检对话框...',
+                'level': 'warning',
+            })
+            critical_items = [it for it in missing_items if it.is_critical]
+            if critical_items:
+                lines = "\n".join(f"  - [{it.step_name}] {it.reason}" for it in critical_items)
+                global_event_bus.publish('LogMessage', {
+                    'message': f'[预检] 阻断性错误 ({len(critical_items)} 项):\n{lines}',
+                    'level': 'error',
+                })
+                QMessageBox.critical(
+                    self.parent(), "预检失败（阻断性错误）",
+                    f"以下为阻断性缺失，流程无法启动：\n\n{lines}\n\n请填写后重新运行。"
+                )
+                return
+
+            dialog = PreflightDialog(missing_items, parent=self.parent())
+            if dialog.exec() != QDialog.Accepted:
+                global_event_bus.publish('LogMessage', {
+                    'message': '⚠ 流程中止：用户取消了预检对话框',
+                    'level': 'warning',
+                })
+                return
+            result = dialog.get_result()
+            if result is None:
+                global_event_bus.publish('LogMessage', {
+                    'message': '⚠ 流程中止：预检对话框返回空结果',
+                    'level': 'warning',
+                })
+                return
+
+            action, *payload = result
+            if action == "fill":
+                _, step_id, tab_index = result
+                global_event_bus.publish('NavigateToTab', {
+                    'tab_index': tab_index,
+                    'step_id': step_id,
+                })
+                global_event_bus.publish('LogMessage', {
+                    'message': f'[预检] 用户选择填写 {step_id}，已切换到对应面板。流程暂停，填写完成后请重新运行。',
+                    'level': 'info',
+                })
+                return
+            skip_list = payload[0] if payload else []
+            if skip_list:
+                global_event_bus.publish('LogMessage', {
+                    'message': f'[预检] 用户强制跳过 {len(skip_list)} 个步骤: {skip_list}',
+                    'level': 'warning',
+                })
+        else:
+            global_event_bus.publish('LogMessage', {
+                'message': '[预检] ✓ 所有必需项均已就绪，无需弹窗',
+                'level': 'info',
+            })
+
+        # ── 8) 确认执行 ──
+        reply = QMessageBox.question(
+            self.parent(), "确认",
+            "是否开始执行完整流程？\n\n这可能需要较长时间，请确保配置正确。",
+            QMessageBox.Yes | QMessageBox.No
+        )
+        if reply != QMessageBox.Yes:
+            global_event_bus.publish('LogMessage', {
+                'message': '⚠ 流程中止：用户取消了执行确认',
+                'level': 'warning',
+            })
+            return
+
+        # ── 9) 准备 worker_config ──
+        global_event_bus.publish('LogMessage', {
+            'message': f'初始化 Pipeline，工作目录: {work_dir}',
+            'level': 'info',
+        })
+
+        worker_config = copy.deepcopy(config)
+        step6_cfg = worker_config.get('step6_feature')
+        if step6_cfg:
+            enabled = step6_cfg.pop('enabled', True)
+            if not enabled:
+                worker_config.pop('step6_feature', None)
+
+        global_event_bus.publish('LogMessage', {
+            'message': f'[运行] 最终执行配置包含 {len(worker_config)} 个步骤: {list(worker_config.keys())}',
+            'level': 'info',
+        })
+
+        # ── 10) 创建 WorkerThread 并连线 ──
+        self._worker = WorkerThread(work_dir, worker_config, mode='full', skip_list=skip_list)
+        self._worker.log_message.connect(self._on_log_message, Qt.QueuedConnection)
+        self._worker.progress_update.connect(self._on_progress_update, Qt.QueuedConnection)
+        self._worker.step_completed.connect(self._on_step_completed, Qt.QueuedConnection)
+        self._worker.finished.connect(self._on_finished, Qt.QueuedConnection)
+
+        # ── 11) 发布启动事件 → 主窗口订阅后禁用按钮 ──
+        global_event_bus.publish('PipelineStarted', {})
+        global_event_bus.publish('ProgressUpdate', {'percentage': 0, 'message': '准备执行...'})
+        global_event_bus.publish('LogMessage', {'message': '=' * 50, 'level': 'info'})
+        global_event_bus.publish('LogMessage', {'message': '开始执行完整流程...', 'level': 'info'})
+        global_event_bus.publish('LogMessage', {'message': '=' * 50, 'level': 'info'})
+
+        self._worker.start()
+
+    def run_single_step(self, step_name: str, config: dict = None):
+        """运行单个步骤。
+
+        Args:
+            step_name: 步骤名称（如 'step1', 'step5_clean'）
+            config: 步骤配置字典（可选，默认从面板获取）
+        """
+        try:
+            self._run_single_step_impl(step_name, config)
+        except Exception as e:
+            err_detail = traceback.format_exc()
+            global_event_bus.publish('LogMessage', {
+                'message': f'[致命错误] run_single_step 异常: {e}',
+                'level': 'error',
+            })
+            global_event_bus.publish('LogMessage', {
+                'message': f'详细追踪:\n{err_detail}',
+                'level': 'error',
+            })
+            QMessageBox.critical(
+                self.parent(), "运行失败",
+                f"启动单步执行时发生未预期的错误:\n\n{e}\n\n详细信息已输出到日志区。"
+            )
+
+    def _run_single_step_impl(self, step_name: str, config: dict = None):
+        if not PIPELINE_AVAILABLE:
+            global_event_bus.publish('LogMessage', {
+                'message': '无法导入 Pipeline 模块，请检查 src/core/handlers/ 目录是否完整！',
+                'level': 'error',
+            })
+            QMessageBox.critical(
+                self.parent(), "错误",
+                "无法导入 Pipeline 模块，请检查 src/core/handlers/ 目录是否完整！"
+            )
+            return
+
+        work_dir = self._workspace_initializer.work_dir or './work_dir'
+
+        if config is None:
+            global_event_bus.publish('LogMessage', {
+                'message': '[运行] 正在收集所有步骤面板的配置...',
+                'level': 'info',
+            })
+            self._panel_factory.preload_all()
+            config = self._get_current_config()
+            global_event_bus.publish('LogMessage', {
+                'message': f'[运行] 已收集 {len(config)} 个步骤的配置',
+                'level': 'info',
+            })
+
+        global_event_bus.publish('LogMessage', {
+            'message': f'初始化 Pipeline，工作目录: {work_dir}',
+            'level': 'info',
+        })
+
+        self._worker = WorkerThread(work_dir, config, mode='single_step', step_name=step_name)
+        self._worker.log_message.connect(self._on_log_message, Qt.QueuedConnection)
+        self._worker.progress_update.connect(self._on_progress_update, Qt.QueuedConnection)
+        self._worker.step_completed.connect(self._on_step_completed, Qt.QueuedConnection)
+        self._worker.finished.connect(self._on_finished, Qt.QueuedConnection)
+
+        global_event_bus.publish('PipelineStarted', {})
+        global_event_bus.publish('ProgressUpdate', {'percentage': 0, 'message': f'准备执行 {step_name}...'})
+        global_event_bus.publish('LogMessage', {'message': '=' * 50, 'level': 'info'})
+        global_event_bus.publish('LogMessage', {
+            'message': f'开始独立运行步骤 {step_name}...',
+            'level': 'info',
+        })
+        global_event_bus.publish('LogMessage', {'message': '=' * 50, 'level': 'info'})
+
+        self._worker.start()
+
+    def stop_pipeline(self):
+        """停止当前执行的流程。"""
+        if self._worker and self._worker.isRunning():
+            reply = QMessageBox.question(
+                self.parent(), "确认",
+                "是否停止当前流程？",
+                QMessageBox.Yes | QMessageBox.No
+            )
+            if reply == QMessageBox.Yes:
+                self._worker.stop()
+                global_event_bus.publish('LogMessage', {
+                    'message': '用户取消执行',
+                    'level': 'warning',
+                })
+                global_event_bus.publish('PipelineStopped', {})
+
+    # ═══════════════════════════════════════════════════════════
+    # EventBus 订阅回调
+    # ═══════════════════════════════════════════════════════════
+
+    def _on_request_run_single_step(self, data: dict):
+        """处理面板通过 EventBus 发出的单步执行请求。
+
+        data 格式: {'step_name': 'step1', 'config': {'step1': {...}}}
+
+        前置条件检查（预检/工作目录）由 run_single_step → _run_single_step_impl
+        内部统一处理，此处仅做解析 + 转发 + 异常兜底。
+        """
+        try:
+            step_name = data.get('step_name')
+            config = data.get('config')
+
+            print(f"==== Executor 收到单步请求: {step_name} ====", flush=True)
+
+            if not step_name:
+                global_event_bus.publish('LogMessage', {
+                    'message': '[单步执行] 请求缺少 step_name，忽略',
+                    'level': 'warning',
+                })
+                return
+
+            # ★ 防死锁：若已有 Worker 在运行，不静默吞掉，而是通知用户
+            if self.is_running:
+                global_event_bus.publish('LogMessage', {
+                    'message': f'[单步执行] 后台正在运行中，无法启动 {step_name}。请等待当前任务完成或手动停止后再试。',
+                    'level': 'warning',
+                })
+                QMessageBox.warning(
+                    self.parent(), "后台忙碌",
+                    f"后台正在运行中，无法启动 {step_name}。\n\n请等待当前任务完成，或点击「停止」按钮后再试。"
+                )
+                return
+
+            global_event_bus.publish('LogMessage', {
+                'message': f'[单步执行] 收到 {step_name} 的执行请求',
+                'level': 'info',
+            })
+
+            self.run_single_step(step_name, config)
+
+        except Exception as e:
+            err_detail = traceback.format_exc()
+            global_event_bus.publish('LogMessage', {
+                'message': f'[致命错误] _on_request_run_single_step({step_name}) 异常: {e}',
+                'level': 'error',
+            })
+            global_event_bus.publish('LogMessage', {
+                'message': f'详细追踪:\n{err_detail}',
+                'level': 'error',
+            })
+
+    # ═══════════════════════════════════════════════════════════
+    # WorkerThread 信号 → EventBus 事件（纯转发，零 UI 操作）
+    # ═══════════════════════════════════════════════════════════
+
+    def _on_log_message(self, message: str, level: str):
+        """WorkerThread 日志 → EventBus LogMessage 事件。"""
+        global_event_bus.publish('LogMessage', {
+            'message': message,
+            'level': level,
+        })
+
+    def _on_progress_update(self, percentage: int, message: str):
+        """WorkerThread 进度 → EventBus ProgressUpdate 事件。"""
+        global_event_bus.publish('ProgressUpdate', {
+            'percentage': percentage,
+            'message': message,
+        })
+
+    def _on_step_completed(self, step_name: str, success: bool, message: str):
+        """WorkerThread 步骤完成 → EventBus StepCompleted 事件。
+
+        WorkspaceInitializer 订阅此事件，自动扫描产物并发布 OutputUpdated。
+        """
+        global_event_bus.publish('StepCompleted', {
+            'step_name': step_name,
+            'success': success,
+            'message': message,
+        })
+
+    def _on_finished(self, success: bool, message: str):
+        """WorkerThread 完成 → EventBus PipelineFinished 事件。
+
+        主窗口订阅此事件，恢复按钮状态并弹窗。
+        """
+        global_event_bus.publish('PipelineFinished', {
+            'success': success,
+            'message': message,
+        })
+
+    # ═══════════════════════════════════════════════════════════
+    # 内部辅助
+    # ═══════════════════════════════════════════════════════════
+
+    def _get_current_config(self) -> dict:
+        """从所有已加载面板收集配置。
+
+        注意：仅收集已加载面板（懒加载模式下可能不全）。
+        如需全量配置，调用方应先执行 panel_factory.preload_all()。
+        """
+        config = {}
+        for step_id, panel in self._panel_factory.get_loaded_panels().items():
+            if hasattr(panel, 'get_config'):
+                config[step_id] = panel.get_config()
+        return config
+
+    def _precheck_step3_bands(self) -> bool:
+        """步骤 3 波段越界预检（主线程同步执行，避免多线程弹窗问题）。
+
+        读取 step1 影像的 RasterCount，校验 step3 面板当前方法下所有波段索引
+        是否越界。若越界，弹 BandConfirmDialog（60s 倒计时）让用户调整或取消。
+
+        Returns:
+            True: 预检通过或已自动调整，继续执行
+            False: 用户点"取消运行"，应中止
+        """
+        try:
+            step1_panel = self._panel_factory.get_panel('step1')
+            step3_panel = self._panel_factory.get_panel('step3')
+            img_path = step1_panel.img_file.get_path() if step1_panel else None
+            step3_cfg = step3_panel.get_config() if step3_panel else None
+            step3_enabled = step3_panel.enable_checkbox.isChecked() if step3_panel else False
+        except Exception as e:
+            global_event_bus.publish('LogMessage', {
+                'message': f'⚠ step3 波段预检：读取面板状态失败 - {e}',
+                'level': 'warning',
+            })
+            return True
+
+        if not step3_enabled:
+            return True
+        if not img_path or not os.path.isfile(img_path):
+            global_event_bus.publish('LogMessage', {
+                'message': '⚠ step3 波段预检：未找到参考影像，跳过',
+                'level': 'info',
+            })
+            return True
+        if not step3_cfg:
+            return True
+
+        try:
+            from osgeo import gdal
+            dataset = gdal.Open(img_path)
+            if dataset is None:
+                global_event_bus.publish('LogMessage', {
+                    'message': f'⚠ step3 波段预检：gdal 无法打开影像 {img_path}',
+                    'level': 'warning',
+                })
+                return True
+            max_band = dataset.RasterCount
+            dataset = None
+        except Exception as e:
+            global_event_bus.publish('LogMessage', {
+                'message': f'⚠ step3 波段预检：读取 RasterCount 失败 - {e}',
+                'level': 'warning',
+            })
+            return True
+
+        if max_band <= 0:
+            return True
+
+        method = step3_cfg.get('method', 'goodman')
+        if method == 'goodman':
+            band_fields = [
+                ('nir_lower', 'nir_lower', 65, 'NIR下波段'),
+                ('nir_upper', 'nir_upper', 91, 'NIR上波段'),
+            ]
+        elif method == 'kutser':
+            band_fields = [
+                ('oxy_band', 'oxy_band', 38, '氧吸收波段'),
+                ('lower_oxy', 'lower_oxy', 36, '下氧吸收波段'),
+                ('upper_oxy', 'upper_oxy', 49, '上氧吸收波段'),
+                ('nir_band', 'nir_band', 47, 'NIR波段'),
+            ]
+        elif method == 'hedley':
+            band_fields = [
+                ('hedley_nir_band', 'hedley_nir_band', 47, 'NIR波段'),
+            ]
+        else:
+            return True
+
+        for cfg_key, panel_attr, recommended, label in band_fields:
+            requested = step3_cfg.get(cfg_key)
+            if requested is None or requested <= max_band:
+                continue
+
+            global_event_bus.publish('LogMessage', {
+                'message': f'⚠ step3 波段越界：{label}={requested} > 影像波段数 {max_band}',
+                'level': 'warning',
+            })
+
+            dlg = BandConfirmDialog(
+                self.parent(),
+                requested_band=requested,
+                max_band=max_band,
+                recommended_band=recommended,
+                method_label=label,
+            )
+            result = dlg.exec_()
+            if result == QDialog.Rejected:
+                global_event_bus.publish('LogMessage', {
+                    'message': '✗ 用户取消运行（step3 波段越界未解决）',
+                    'level': 'warning',
+                })
+                return False
+
+            new_band = dlg.selected_band()
+            try:
+                spin = getattr(step3_panel, panel_attr)
+                spin.setValue(new_band)
+            except AttributeError:
+                global_event_bus.publish('LogMessage', {
+                    'message': f'⚠ step3 panel 缺控件 {panel_attr}，跳过回写',
+                    'level': 'warning',
+                })
+                continue
+
+            global_event_bus.publish('LogMessage', {
+                'message': f'✓ {label}：{requested} → {new_band}（影像最多 {max_band} 波段）',
+                'level': 'info',
+            })
+
+        return True

src/gui/core/pipeline_mode_dialog.py

@@ -0,0 +1,237 @@
+# -*- coding: utf-8 -*-
+"""
+PipelineModeDialog：全流程运行前的模式选择弹窗。
+
+用户点击"运行完整流程"后，首先弹出此弹窗选择执行模式：
+  - 选项 A（训练新模型并预测）：执行完整建模与预测流程，需要实测水质 CSV
+  - 选项 B（使用已有模型直接预测）：跳过训练步骤，直接使用外部模型目录进行预测
+
+弹窗结果：
+  - QDialog.Accepted + self.selected_mode = "training" 或 "prediction_only"
+  - QDialog.Rejected → 调用方中止 run_full_pipeline
+"""
+
+import os
+from typing import Optional
+
+from PyQt5.QtCore import Qt
+from PyQt5.QtGui import QFont
+from PyQt5.QtWidgets import (
+    QDialog, QVBoxLayout, QHBoxLayout, QLabel, QPushButton,
+    QRadioButton, QGroupBox, QButtonGroup, QMessageBox, QSizePolicy,
+)
+
+
+def _is_valid_model_dir(path: str) -> bool:
+    """深层递归检测模型目录：只要任意层级存在文件即返回 True。"""
+    if not path or not os.path.isdir(path):
+        return False
+    for _root, _dirs, files in os.walk(path):
+        if files:
+            return True
+    return False
+
+
+class PipelineModeDialog(QDialog):
+    """全流程模式选择对话框。
+
+    两个单选按钮覆盖两种业务场景：
+      - A：训练新模型（完整流程，需要 step4 CSV）
+      - B：仅预测（跳过 step4/5/7/8，直接用外部模型目录）
+
+    属性：
+      selected_mode: "training" | "prediction_only"
+    """
+
+    def __init__(self, main_window=None, parent=None):
+        super().__init__(parent)
+        self.main_window = main_window
+        self.selected_mode: Optional[str] = None
+        self.setWindowTitle("选择运行模式")
+        self.setMinimumSize(560, 340)
+        self.setModal(True)
+        self._setup_ui()
+
+    # ------------------------------------------------------------------
+    # UI 构建
+    # ------------------------------------------------------------------
+
+    def _setup_ui(self):
+        layout = QVBoxLayout(self)
+        layout.setContentsMargins(28, 24, 28, 20)
+        layout.setSpacing(14)
+
+        # ── 标题 ──
+        title = QLabel("请选择全流程运行模式")
+        title_font = QFont()
+        title_font.setPointSize(13)
+        title_font.setBold(True)
+        title.setFont(title_font)
+        title.setAlignment(Qt.AlignCenter)
+        layout.addWidget(title)
+
+        layout.addSpacing(4)
+
+        # ── 选项 A：训练新模型 ──
+        group_a = QGroupBox()
+        group_a.setObjectName("groupA")
+        group_a.setMinimumHeight(100)
+        layout.addWidget(group_a)
+
+        self.radio_a = QRadioButton("【训练新模型并预测】")
+        self.radio_a.setChecked(True)  # 默认选项 A
+        self.radio_a.setObjectName("radioTraining")
+
+        desc_a = QLabel(
+            "需要提供实测水质数据 (CSV)，将执行完整建模与预测流程。\n"
+            "包括：水域掩膜 → 耀斑去除 → 光谱特征提取 → 模型训练 → 密集采样 → 预测 → 专题图"
+        )
+        desc_a.setWordWrap(True)
+        desc_a.setStyleSheet("color: #555555; background: transparent;")
+        desc_a.setObjectName("descA")
+
+        vbox_a = QVBoxLayout(group_a)
+        vbox_a.setContentsMargins(16, 20, 16, 14)
+        vbox_a.setSpacing(8)
+        vbox_a.addWidget(self.radio_a)
+        vbox_a.addWidget(desc_a)
+
+        # ── 选项 B：仅预测 ──
+        group_b = QGroupBox()
+        group_b.setObjectName("groupB")
+        group_b.setMinimumHeight(100)
+        layout.addWidget(group_b)
+
+        self.radio_b = QRadioButton("【使用已有模型直接预测】")
+        self.radio_b.setObjectName("radioPrediction")
+
+        desc_b = QLabel(
+            "跳过模型训练步骤，直接使用导入的外部模型目录进行预测。\n"
+            "前提条件：请在「监督预测」或「回归预测」面板中指定模型目录。\n"
+            "适用范围：已有预训练模型、或其他来源模型目录。"
+        )
+        desc_b.setWordWrap(True)
+        desc_b.setStyleSheet("color: #555555; background: transparent;")
+        desc_b.setObjectName("descB")
+
+        vbox_b = QVBoxLayout(group_b)
+        vbox_b.setContentsMargins(16, 20, 16, 14)
+        vbox_b.setSpacing(8)
+        vbox_b.addWidget(self.radio_b)
+        vbox_b.addWidget(desc_b)
+
+        # ── 强制互斥：QButtonGroup ──
+        self.mode_group = QButtonGroup(self)
+        self.mode_group.addButton(self.radio_a)
+        self.mode_group.addButton(self.radio_b)
+
+        # ── 提示栏（动态显示 models_dir 状态） ──
+        self.models_hint = QLabel()
+        self.models_hint.setObjectName("modelsHint")
+        self.models_hint.setWordWrap(True)
+        self.models_hint.setStyleSheet("color: #888888; font-size: 11px; padding: 4px 0;")
+        layout.addWidget(self.models_hint)
+
+        # ── 强制 QRadioButton 指示器为实心圆点 ──
+        self.setStyleSheet("""
+            QRadioButton::indicator {
+                width: 14px;
+                height: 14px;
+            }
+            QRadioButton::indicator:checked {
+                background-color: #0078D7;
+                border: 2px solid #0078D7;
+                border-radius: 7px;
+            }
+            QRadioButton::indicator:unchecked {
+                background-color: white;
+                border: 2px solid #A0A0A0;
+                border-radius: 7px;
+            }
+        """)
+
+        # ── 按钮 ──
+        btn_layout = QHBoxLayout()
+        btn_layout.addStretch()
+
+        cancel_btn = QPushButton("取消")
+        cancel_btn.setObjectName("cancelBtn")
+        cancel_btn.setMinimumWidth(90)
+        cancel_btn.clicked.connect(self.reject)
+
+        self.btn_confirm = QPushButton("确认")
+        self.btn_confirm.setObjectName("confirmBtn")
+        self.btn_confirm.setMinimumWidth(90)
+        self.btn_confirm.setDefault(True)
+        self.btn_confirm.clicked.connect(self._on_confirm)
+
+        btn_layout.addWidget(self.btn_confirm)
+        btn_layout.addWidget(cancel_btn)
+        layout.addLayout(btn_layout)
+
+        # 信号连接：任一 radio 切换时重新渲染提示 + 按钮状态
+        self.radio_a.toggled.connect(self._update_models_hint)
+        self.radio_b.toggled.connect(self._update_models_hint)
+
+        # 初始状态渲染
+        self._update_models_hint()
+
+    def _update_models_hint(self, checked=False, *args) -> None:
+        """根据当前选中模式和 models_dir 状态更新提示文字及确认按钮可用性。"""
+        training_checked = self.radio_a.isChecked()
+
+        # 从主窗口 config 读取 models_dir（优先 ml，其次 reg）
+        models_dir = ""
+        if self.main_window:
+            config = self.main_window.get_current_config()
+            models_dir = config.get("step11_ml", {}).get("models_dir", "")
+            if not models_dir:
+                models_dir = config.get("step11", {}).get("models_dir", "")
+
+        has_files = bool(models_dir and _is_valid_model_dir(models_dir))
+        dir_exists = bool(models_dir and os.path.isdir(models_dir))
+
+        if training_checked:
+            if hasattr(self, 'btn_confirm') and self.btn_confirm is not None:
+                self.btn_confirm.setEnabled(True)
+            if has_files:
+                self.models_hint.setText(
+                    f"⚠ 注意：当前模型目录已包含文件，继续训练将会【覆盖】原有模型！\n路径：{models_dir}"
+                )
+                self.models_hint.setStyleSheet("color: #e65100; font-size: 11px; padding: 4px 0;")
+            else:
+                label = f"✓ 模型将保存至该目录（当前为空，安全）。\n路径：{models_dir}" if dir_exists else "✓ 尚未指定模型目录，将使用默认路径创建新模型。"
+                self.models_hint.setText(label)
+                self.models_hint.setStyleSheet("color: #2e7d32; font-size: 11px; padding: 4px 0;")
+        else:
+            if has_files:
+                self.models_hint.setText(
+                    f"✓ 已检测到有效模型目录，可以直接预测。\n路径：{models_dir}"
+                )
+                self.models_hint.setStyleSheet("color: #2e7d32; font-size: 11px; padding: 4px 0;")
+                if hasattr(self, 'btn_confirm') and self.btn_confirm is not None:
+                    self.btn_confirm.setEnabled(True)
+            else:
+                if dir_exists:
+                    self.models_hint.setText(
+                        f"❌ 错误：模型目录为空（未找到任何文件），无法进行预测！\n路径：{models_dir}"
+                    )
+                else:
+                    self.models_hint.setText(
+                        "❌ 错误：模型目录为空或不存在！请先返回对应面板配置有效路径。"
+                    )
+                self.models_hint.setStyleSheet("color: #c62828; font-size: 11px; padding: 4px 0;")
+                if hasattr(self, 'btn_confirm') and self.btn_confirm is not None:
+                    self.btn_confirm.setEnabled(False)
+
+    def _on_confirm(self) -> None:
+        """确认按钮回调：直接存储模式并关闭。
+
+        注意：按钮禁用状态已在 _update_models_hint 中处理，
+        此处仅负责结果存储，不再做二次弹窗拦截。
+        """
+        if self.radio_a.isChecked():
+            self.selected_mode = "training"
+        else:
+            self.selected_mode = "prediction_only"
+        self.accept()

src/gui/core/preflight_dialog.py

@@ -0,0 +1,429 @@
+# -*- coding: utf-8 -*-
+"""
+预检交互对话框：一次性全预检 + 用户交互式决策。
+
+用户点击"运行"后，若存在缺失项：
+  - 列出每个缺失项（步骤名 + 原因）
+  - 每项提供"填写"（跳转面板）和"忽略"（加入 skip_list）选项
+  - 底部三个操作按钮决定流程走向
+"""
+import os
+from dataclasses import dataclass
+from typing import Dict, List, Optional, Set, Tuple
+
+from PyQt5.QtWidgets import (
+    QDialog, QVBoxLayout, QHBoxLayout, QLabel, QPushButton,
+    QScrollArea, QWidget, QCheckBox, QGroupBox, QFrame,
+    QSizePolicy, QStyleFactory,
+)
+from PyQt5.QtCore import Qt
+from PyQt5.QtGui import QFont, QColor, QPalette
+
+from src.core.pipeline.runner import PIPELINE_STEPS
+
+
+@dataclass
+class MissingItem:
+    """单个缺失项的结构化描述"""
+    step_id: str          # step_id，如 "step1"、"step8_non_empirical_modeling"
+    step_name: str        # 面板 tab 显示名称，如 "水域掩膜"
+    reason: str           # 缺失原因，如 "缺少参考影像路径"
+    panel_tab_index: int  # step_stack 中的 tab 索引（用于切换）
+    is_critical: bool = False  # 是否为阻断性缺失（img_path 缺失 = True）
+
+
+# ============================================================
+# PreflightDialog
+# ============================================================
+
+class PreflightDialog(QDialog):
+    """预检交互对话框。
+
+    对每个 MissingItem，用户可选择：
+      - 勾选"忽略"：将该 step_id 加入 skip_list，运行时跳过
+      - 点击"填写"：关闭弹窗，切换到对应面板 tab
+
+    对话框结果 (exec 返回值)：
+      - QDialog.Accepted + self.result_data = ("fill", step_id)
+        → 填写待办：切换到目标面板，停止流程
+      - QDialog.Accepted + self.result_data = ("skip", skip_list)
+        → 强制跳过：携带 skip_list 继续运行
+      - QDialog.Rejected
+        → 取消运行：完全停止
+    """
+
+    # step_id → (step_name, panel_tab_index)
+    STEP_TAB_MAP = {
+        "step1":   ("水域掩膜",     0),
+        "step2":   ("耀斑检测",     1),
+        "step3":   ("耀斑去除",     2),
+        "step4":   ("数据清洗",     3),
+        "step5":   ("特征构建",     4),
+        "step7":   ("水质指数",     5),
+        "step8_non_empirical_modeling": ("回归建模", 7),
+        "step9":   ("水色指数反演",  8),
+        "step10":  ("采样点布设",   10),
+        "step11_ml": ("监督预测",  11),
+        "step11":  ("回归预测",    12),
+        "step14":  ("专题图生成",  13),
+    }
+
+    def __init__(self, missing_items: List[MissingItem], parent=None):
+        super().__init__(parent)
+        self.missing_items = missing_items
+        self.result_data: Optional[Tuple[str, any]] = None  # ("fill", step_id) | ("skip", [step_id])
+        self._skip_checkboxes: List[QCheckBox] = []
+        self._fill_buttons: List[QPushButton] = []
+
+        self.setWindowTitle("⚠ 预检发现缺失项")
+        self.setMinimumSize(680, 420)
+        self.setModal(True)
+        self._setup_ui()
+
+    # ------------------------------------------------------------------
+    # UI 构建
+    # ------------------------------------------------------------------
+
+    def _setup_ui(self):
+        main_layout = QVBoxLayout(self)
+        main_layout.setContentsMargins(20, 20, 20, 16)
+        main_layout.setSpacing(10)
+
+        # ── 顶部提示 ──
+        header_label = QLabel(
+            f"检测到 <b>{len(self.missing_items)}</b> 个缺失项，请逐项处理后继续："
+        )
+        header_label.setStyleSheet("font-size: 14px; color: #e67e22; font-weight: bold;")
+        main_layout.addWidget(header_label)
+
+        # ── 滚动区域（缺失项列表） ──
+        scroll = QScrollArea()
+        scroll.setWidgetResizable(True)
+        scroll.setFrameShape(QFrame.NoFrame)
+        scroll.setStyleSheet("background: transparent;")
+
+        container = QWidget()
+        container_layout = QVBoxLayout(container)
+        container_layout.setContentsMargins(0, 0, 8, 0)
+        container_layout.setSpacing(8)
+
+        for item in self.missing_items:
+            row = self._build_item_row(item)
+            container_layout.addWidget(row)
+
+        container_layout.addStretch()
+        scroll.setWidget(container)
+        main_layout.addWidget(scroll, 1)
+
+        # ── 底部操作按钮 ──
+        btn_layout = QHBoxLayout()
+        btn_layout.setSpacing(12)
+
+        # 取消运行（左）
+        cancel_btn = QPushButton("取消运行")
+        cancel_btn.setCursor(Qt.PointingHandCursor)
+        cancel_btn.setMinimumHeight(38)
+        cancel_btn.setStyleSheet(
+            "QPushButton { background: #95a5a6; color: white; border-radius: 6px; "
+            "font-weight: bold; font-size: 13px; padding: 4px 16px; }"
+            "QPushButton:hover { background: #7f8c8d; }"
+        )
+        cancel_btn.clicked.connect(self._on_cancel)
+        btn_layout.addWidget(cancel_btn)
+
+        btn_layout.addStretch()
+
+        # 强制跳过运行（中）
+        skip_btn = QPushButton("强制跳过运行")
+        skip_btn.setCursor(Qt.PointingHandCursor)
+        skip_btn.setMinimumHeight(38)
+        skip_btn.setStyleSheet(
+            "QPushButton { background: #3498db; color: white; border-radius: 6px; "
+            "font-weight: bold; font-size: 13px; padding: 4px 16px; }"
+            "QPushButton:hover { background: #2980b9; }"
+        )
+        skip_btn.clicked.connect(self._on_force_skip)
+        btn_layout.addWidget(skip_btn)
+
+        # 填写待办（右，primary）
+        fill_btn = QPushButton("填写待办")
+        fill_btn.setCursor(Qt.PointingHandCursor)
+        fill_btn.setMinimumHeight(38)
+        fill_btn.setDefault(True)
+        fill_btn.setAutoDefault(True)
+        fill_btn.setStyleSheet(
+            "QPushButton { background: #27ae60; color: white; border-radius: 6px; "
+            "font-weight: bold; font-size: 13px; padding: 4px 16px; }"
+            "QPushButton:hover { background: #1e8449; }"
+        )
+        fill_btn.clicked.connect(self._on_fill_first)
+        btn_layout.addWidget(fill_btn)
+
+        main_layout.addLayout(btn_layout)
+
+    def _build_item_row(self, item: MissingItem) -> QWidget:
+        """构建单个缺失项行 widget。"""
+        frame = QFrame()
+        frame.setFrameShape(QFrame.StyledPanel)
+        frame.setStyleSheet(
+            "QFrame { background: #2c3e50; border-radius: 8px; padding: 10px; }"
+            "QFrame[critical=true] { border: 2px solid #e74c3c; }"
+            "QFrame[critical=false] { border: 1px solid #34495e; }"
+        )
+        frame.setProperty("critical", item.is_critical)
+
+        layout = QVBoxLayout(frame)
+        layout.setContentsMargins(12, 10, 12, 10)
+        layout.setSpacing(6)
+
+        # ── 第一行：步骤标签 + 原因 ──
+        top = QHBoxLayout()
+        top.setSpacing(8)
+
+        # 步骤名标签
+        name_label = QLabel(f"📌 {item.step_name}")
+        name_label.setStyleSheet(
+            "font-size: 13px; font-weight: bold; color: #f39c12; background: #1a252f; "
+            "border-radius: 4px; padding: 4px 10px;"
+        )
+        top.addWidget(name_label)
+
+        # 阻断性标记
+        if item.is_critical:
+            critical_label = QLabel("阻断")
+            critical_label.setStyleSheet(
+                "background: #e74c3c; color: white; border-radius: 4px; "
+                "font-size: 11px; font-weight: bold; padding: 3px 8px;"
+            )
+            top.addWidget(critical_label)
+
+        top.addStretch()
+
+        # "填写"按钮
+        fill_btn = QPushButton("填写")
+        fill_btn.setCursor(Qt.PointingHandCursor)
+        fill_btn.setFixedWidth(70)
+        fill_btn.setFixedHeight(28)
+        fill_btn.setStyleSheet(
+            "QPushButton { background: #27ae60; color: white; border-radius: 5px; "
+            "font-size: 12px; font-weight: bold; }"
+            "QPushButton:hover { background: #1e8449; }"
+        )
+        fill_btn.clicked.connect(lambda *a, sid=item.step_id, idx=item.panel_tab_index: self._on_fill(sid, idx))
+        self._fill_buttons.append(fill_btn)
+        top.addWidget(fill_btn)
+
+        layout.addLayout(top)
+
+        # ── 第二行：原因文本 ──
+        reason_label = QLabel(item.reason)
+        reason_label.setWordWrap(True)
+        reason_label.setStyleSheet(
+            "font-size: 12px; color: #bdc3c7; background: transparent; padding: 2px 4px;"
+        )
+        reason_label.setTextInteractionFlags(Qt.TextSelectableByMouse)
+        layout.addWidget(reason_label)
+
+        # ── 第三行：忽略复选框 ──
+        bottom = QHBoxLayout()
+        bottom.addStretch()
+
+        skip_cb = QCheckBox("忽略此项（强制跳过）")
+        skip_cb.setCursor(Qt.PointingHandCursor)
+        skip_cb.setStyleSheet(
+            "QCheckBox { color: #95a5a6; font-size: 12px; spacing: 6px; }"
+            "QCheckBox::indicator { width: 16px; height: 16px; }"
+        )
+        skip_cb.setChecked(False)
+        skip_cb.stateChanged.connect(
+            lambda state, cb=skip_cb: cb.setStyleSheet(
+                "QCheckBox { color: #27ae60; font-size: 12px; spacing: 6px; }"
+                "QCheckBox::indicator { width: 16px; height: 16px; }"
+                if state else
+                "QCheckBox { color: #95a5a6; font-size: 12px; spacing: 6px; }"
+                "QCheckBox::indicator { width: 16px; height: 16px; }"
+            )
+        )
+        self._skip_checkboxes.append((item.step_id, skip_cb))
+        bottom.addWidget(skip_cb)
+
+        layout.addLayout(bottom)
+        return frame
+
+    # ------------------------------------------------------------------
+    # 槽函数
+    # ------------------------------------------------------------------
+
+    def _on_cancel(self):
+        """取消运行：完全停止。"""
+        self.result_data = None
+        self.reject()
+
+    def _on_force_skip(self):
+        """强制跳过：收集所有被勾选"忽略"的 step_id，携带 skip_list 继续。"""
+        skip_list = [
+            step_id for step_id, cb in self._skip_checkboxes if cb.isChecked()
+        ]
+        self.result_data = ("skip", skip_list)
+        self.accept()
+
+    def _on_fill_first(self):
+        """填写待办：找到第一个未被勾选"忽略"的缺失项，切换到其面板。"""
+        for step_id, cb in self._skip_checkboxes:
+            if not cb.isChecked():
+                item = self._find_item(step_id)
+                if item:
+                    self.result_data = ("fill", step_id, item.panel_tab_index)
+                    self.accept()
+                    return
+        # 所有项都被勾选 → 等同于 force_skip
+        self._on_force_skip()
+
+    def _on_fill(self, step_id: str, tab_index: int):
+        """填写：直接切换到指定面板。"""
+        self.result_data = ("fill", step_id, tab_index)
+        self.accept()
+
+    # ------------------------------------------------------------------
+    # 辅助
+    # ------------------------------------------------------------------
+
+    def _find_item(self, step_id: str) -> Optional[MissingItem]:
+        for item in self.missing_items:
+            if item.step_id == step_id:
+                return item
+        return None
+
+    def get_result(self) -> Optional[Tuple[str, any]]:
+        """供外部获取结果。"""
+        return self.result_data
+
+    @staticmethod
+    def build_missing_items(config: dict) -> List[MissingItem]:
+        """DAG-aware 预检：从 config 构建缺失项列表。
+
+        拓扑预判逻辑：
+        1. 按 pipeline 顺序遍历所有 enabled=True 的步骤，收集其 produces 列表，
+           构建「动态产物集合」dynamically_produced_keys。
+        2. 检查某个 required_input_file 时：
+           - 若磁盘已存在 → OK（用户已手动提供）
+           - 若 key 在 dynamically_produced_keys 中 → OK（前置步骤会生成）
+           - 否则 → MissingItem（真正缺失）
+        3. 智能免检规则：
+           - formula_csv_path：底层的完全可选参数，任何情况下都免检。
+           - step5 boundary_path：若 step1 enabled 或 config 中有 water_mask_path，
+             则信任 panel/底层的自动推导机制，不拦截。
+           - step14 boundary_shp_path：若 step1 enabled，信任 panel 的自动回填，
+             不拦截。
+
+        关键阻断项（is_critical=True）：step1 img_path 缺失。
+        """
+        items: List[MissingItem] = []
+
+        step1_cfg = config.get('step1', {})
+        step1_enabled = step1_cfg.get('enabled', False)
+
+        # ── ★ 构建「动态产物集合」：按 pipeline 顺序收集所有 enabled 步骤的 produces ──
+        dynamically_produced_keys: Set[str] = set()
+        enabled_step_ids: Set[str] = set()
+
+        for step_spec in PIPELINE_STEPS:
+            step_cfg = config.get(step_spec.step_id, {})
+            if not step_cfg.get('enabled', True):
+                continue
+            enabled_step_ids.add(step_spec.step_id)
+            dynamically_produced_keys.update(step_spec.produces)
+
+        # ── step1 img_path（阻断性）───────────────────────────────
+        img_path = step1_cfg.get('img_path')
+        if not img_path:
+            items.append(MissingItem(
+                step_id="step1", step_name="水域掩膜",
+                reason="缺少参考影像路径 → 请在「阶段一」中填写「参考影像」",
+                panel_tab_index=0, is_critical=True
+            ))
+        elif not os.path.isfile(img_path):
+            items.append(MissingItem(
+                step_id="step1", step_name="水域掩膜",
+                reason=f"参考影像文件不存在：{img_path}",
+                panel_tab_index=0, is_critical=True
+            ))
+
+        # ── step4 csv_path（纯外部输入，必须手动提供）───────────────
+        step4_cfg = config.get('step4', {})
+        step4_enabled = step4_cfg.get('enabled', True)
+        if step4_enabled:
+            csv_path = step4_cfg.get('csv_path')
+            if not csv_path:
+                items.append(MissingItem(
+                    step_id="step4", step_name="数据清洗",
+                    reason="请在「数据清洗」中填写「实测水质数据 CSV」",
+                    panel_tab_index=3
+                ))
+            elif not os.path.isfile(csv_path):
+                items.append(MissingItem(
+                    step_id="step4", step_name="数据清洗",
+                    reason=f"实测水质数据文件不存在：{csv_path}",
+                    panel_tab_index=3
+                ))
+
+        # ── step12 formula_csv_path（绝对免检：底层完全可选）────────
+        # formula_csv_path 在底层 CustomRegressionPredictor 中不传即可运行，
+        # 只影响日志输出，不阻断任何功能。此处不做任何检查。
+
+        # ── ★ DAG-aware 检查：遍历 enabled 步骤的 required_input_files ──
+        PURE_EXTERNAL_INPUT_KEYS: Set[str] = {'img_path', 'csv_path'}
+        _TAB_INDEX_MAP: Dict[str, int] = {
+            "step1": 0, "step2": 1, "step3": 2, "step4": 3,
+            "step5": 4, "step8": 5, "step7": 6,
+            "step8_non_empirical_modeling": 7, "step9": 8,
+            "step10": 9, "step11_ml": 10, "step11": 11,
+            "step12": 12, "step14": 13,
+        }
+        _STEP_NAME_MAP: Dict[str, str] = {
+            "step1": "水域掩膜", "step2": "耀斑检测", "step3": "耀斑去除",
+            "step4": "数据清洗", "step5": "特征构建", "step8": "水质指数",
+            "step7": "监督建模", "step8_non_empirical_modeling": "回归建模",
+            "step9": "自定义回归建模", "step10": "采样点布设",
+            "step11_ml": "监督预测", "step11": "回归预测",
+            "step12": "自定义回归预测", "step14": "专题图生成",
+        }
+
+        for step_spec in PIPELINE_STEPS:
+            if step_spec.step_id not in enabled_step_ids:
+                continue
+            step_cfg = config.get(step_spec.step_id, {})
+            tab_idx = _TAB_INDEX_MAP.get(step_spec.step_id, 0)
+            step_name = _STEP_NAME_MAP.get(step_spec.step_id, step_spec.step_id)
+
+            for req_key in step_spec.required_input_files:
+                # ★★★ 高优先级硬编码白名单 ★★★
+                # 当检测到需求为边界文件时，只要 step1 有填影像（代表有基础，底层能自动推导），直接放行
+                if req_key in ('boundary_path', 'boundary_shp_path'):
+                    step1_cfg = config.get('step1', {})
+                    if step1_cfg.get('img_path') or step1_cfg.get('enabled', True):
+                        continue  # 直接跳过，不判定为缺失
+                if req_key in PURE_EXTERNAL_INPUT_KEYS:
+                    continue
+                if req_key == 'formula_csv_path':
+                    continue  # ★ 底层完全可选，赦免
+                if req_key == 'boundary_path' and step_spec.step_id == 'step5':
+                    continue  # ★ step1 执行则 panel/底层自动推导，赦免
+                if req_key == 'boundary_shp_path' and step_spec.step_id == 'step14':
+                    continue  # ★ step1 执行则 panel 自动回填，赦免
+                cfg_val = step_cfg.get(req_key)
+                if cfg_val and os.path.isfile(cfg_val):
+                    continue
+                if cfg_val and os.path.isdir(cfg_val):
+                    continue
+                if req_key in dynamically_produced_keys:
+                    continue  # ★ 前置步骤会生成，拓扑预判通过
+                items.append(MissingItem(
+                    step_id=step_spec.step_id,
+                    step_name=step_name,
+                    reason=f"缺少必需文件/目录 [{req_key}]",
+                    panel_tab_index=tab_idx,
+                    is_critical=(step_spec.step_id == "step1" and req_key == "img_path"),
+                ))
+
+        return items