feat(step10): 新增 WaterIndexCsvProcessor 散点处理入口

This commit is contained in:
DXC
2026-06-24 11:39:52 +08:00
parent c5a82ec342
commit b0ace0bde8

View File

@ -644,3 +644,190 @@ class WaterIndexProcessor:
notify("水色指数反演完成", 100)
return results
# ------------------------------------------------------------------
# 散点处理入口(Step 10 重构后使用,与 Step 9 对称)
# ------------------------------------------------------------------
class WaterIndexCsvProcessor:
"""
散点 CSV 驱动的水色指数反演器。
设计目的
--------
与 Step 9 (ML 预测) 完全对称的【散点处理模式】:
* 输入:Step 4 生成的 ``sampling_spectra.csv``,列结构为
``x_coord, y_coord, pixel_x, pixel_y, 400.000000, 401.000000, ...``
* 处理:解析 ``waterindex.csv`` 中的公式,对每行采样点
提取对应波段数值、逐行 eval 计算水色指数
* 输出:每个公式一个 CSV,列严格为 ``longitude, latitude, <formula_name>``,
可直接喂给 Step 11 ContentMapper
输出目录
--------
默认 ``{work_dir}/10_WaterIndex_CSV/``;若用户指定 ``output_dir`` 则用其值。
"""
COORD_RENAME_MAP = {
"x_coord": "longitude",
"y_coord": "latitude",
"lon": "longitude",
"lat": "latitude",
}
def __init__(self, waterindex_csv_path: Optional[str] = None):
if waterindex_csv_path is None:
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):
waterindex_csv_path = p
break
self.waterindex_csv_path = waterindex_csv_path
self._index_calc = None
def _get_index_calc(self):
"""懒加载 WaterQualityIndexCalculator(首次访问时实例化)"""
if self._index_calc is None and self.waterindex_csv_path:
from src.utils.water_index import WaterQualityIndexCalculator
self._index_calc = WaterQualityIndexCalculator(self.waterindex_csv_path)
return self._index_calc
@staticmethod
def _detect_wavelength_columns(df: "pd.DataFrame") -> List[str]:
"""识别光谱列:列名是浮点数字符串(Step 4 输出的 '400.000000' 形式)"""
import re
wl_cols = []
for col in df.columns:
try:
float(str(col).strip())
wl_cols.append(col)
except (ValueError, TypeError):
continue
return wl_cols
@staticmethod
def _safe_filename(name: str) -> str:
"""公式名 → 文件名安全字符(与旧 BSQ 输出命名习惯一致)"""
return re.sub(r'[^\w\u4e00-\u9fff-]', '_', name).strip('_') or 'index'
def compute_indices_from_csv(
self,
sampling_csv_path: str,
output_dir: str,
selected_formulas: Optional[List[str]] = None,
progress_callback: Optional[Callable[[str, float], None]] = None,
) -> Dict[str, str]:
"""
散点 CSV → 按指数拆分的多个 CSV。
Parameters
----------
sampling_csv_path : str
Step 4 输出的 ``sampling_spectra.csv`` 路径
output_dir : str
输出目录;不存在会自动创建
selected_formulas : list, optional
要计算的公式名列表;None 或空列表 = 全部公式
progress_callback : callable, optional
进度回调 ``(msg: str, pct: float)``
Returns
-------
dict
``{公式名: 输出 CSV 路径}``
"""
def notify(msg: str, pct: float) -> None:
if progress_callback:
progress_callback(msg, pct)
if not os.path.isfile(sampling_csv_path):
raise FileNotFoundError(f"采样点 CSV 不存在: {sampling_csv_path}")
if not self.waterindex_csv_path or not os.path.isfile(self.waterindex_csv_path):
raise FileNotFoundError(
f"waterindex.csv 未配置或不存在: {self.waterindex_csv_path}"
)
os.makedirs(output_dir, exist_ok=True)
notify("正在读取采样点 CSV…", 5)
import pandas as pd
df = pd.read_csv(sampling_csv_path, encoding="utf-8-sig")
if df.empty:
raise ValueError(f"采样点 CSV 为空: {sampling_csv_path}")
# 坐标列重命名(x_coord → longitude, y_coord → latitude)
df = df.rename(columns={k: v for k, v in self.COORD_RENAME_MAP.items()
if k in df.columns})
if "longitude" not in df.columns or "latitude" not in df.columns:
raise ValueError(
f"采样点 CSV 缺少坐标列(期望 x_coord/y_coord 或 longitude/latitude),"
f"实际列: {list(df.columns)}"
)
# 识别光谱列
wl_cols = self._detect_wavelength_columns(df)
if not wl_cols:
raise ValueError(
f"采样点 CSV 中未识别到任何光谱列(列名为数字),"
f"实际列: {list(df.columns)}"
)
notify(f"识别到 {len(wl_cols)} 个光谱列, 采样点 {len(df)} 个", 15)
calc = self._get_index_calc()
if calc is None:
raise RuntimeError("WaterQualityIndexCalculator 初始化失败")
all_formula_names = calc.list_available()
if selected_formulas:
targets = [n for n in selected_formulas if n in all_formula_names]
missing = [n for n in selected_formulas if n not in all_formula_names]
if missing:
print(f"[WaterIndexCsvProcessor] 警告: 以下公式未在 waterindex.csv 中找到,已跳过: {missing}")
else:
targets = all_formula_names
if not targets:
raise ValueError("没有可计算的公式(selected_formulas 为空且 waterindex.csv 中无公式)")
# 一次性算出所有目标公式的 Series(避免重复遍历 DataFrame)
notify(f"开始逐行计算 {len(targets)} 个公式…", 25)
spectra_df = df[wl_cols]
try:
results_df = calc.calculate_many(targets, spectra_df)
except Exception as e:
raise RuntimeError(f"公式计算失败: {e}")
# 每个公式一个 CSV:longitude, latitude, <formula_name>
out_files: Dict[str, str] = {}
n_total = len(targets)
for i, name in enumerate(targets):
try:
per_idx = results_df[name]
out_df = pd.DataFrame({
"longitude": df["longitude"].values,
"latitude": df["latitude"].values,
name: per_idx.values,
})
out_path = os.path.join(output_dir, f"{self._safe_filename(name)}.csv")
out_df.to_csv(out_path, index=False, float_format="%.6f", encoding="utf-8-sig")
out_files[name] = out_path
notify(
f"[{i + 1}/{n_total}] {name} → {os.path.basename(out_path)}",
25 + 70 * (i + 1) / n_total,
)
except Exception as e:
print(f"[WaterIndexCsvProcessor] 公式 '{name}' 失败: {e}")
continue
notify(f"完成!共输出 {len(out_files)} / {n_total} 个指数 CSV", 100)
return out_files
# 保留旧 import 路径兼容
__all__ = ['WaterIndexProcessor', 'WaterIndexCsvProcessor']