refactor: 抽取 _preprocess_dual_stream 独立方法

- preprocess_spectra 中 DualStream_MNF 分支从 25 行内联代码
  变为 1 行方法调用: self._preprocess_dual_stream(spectra, metadata)
- 新方法含完整文档、fallback 路径、重采样逻辑
- 与旧 WQI 补齐/linspace 路径完全解耦

src/core/prediction/inference_batch.py

@@ -433,6 +433,47 @@ class WaterQualityInference:
 
         return None
 
+    # ═══════════════════════════════════════════════════════════
+    # DualStream_MNF 专用预处理：纯光谱重采样 → Pipeline 全自动
+    # ═══════════════════════════════════════════════════════════
+
+    def _preprocess_dual_stream(self, spectra: pd.DataFrame,
+                                metadata: dict) -> np.ndarray:
+        """纯光谱重采样到训练波长网格，返回后由 pipeline.predict() 全自动处理。
+
+        Pipeline 内置 FeatureUnion[PhysicalExtractor + MNFTransformer]，
+        会自动完成物理指数计算和 MNF 降维，无需外部补 WQI。
+        """
+        train_wl = metadata.get('train_wavelengths', None)
+        if train_wl is None or len(train_wl) == 0:
+            print("[DualStream_MNF] ⚠ 模型无 train_wavelengths，fallback 原样输入")
+            return spectra.values
+
+        # 提取纯光谱列
+        spec_cols = []
+        for c in spectra.columns:
+            try:
+                float(str(c)); spec_cols.append(c)
+            except (ValueError, TypeError):
+                pass
+
+        # np.interp 重采样：308/113/任意波段 → 模型训练波长
+        spec_data = spectra[spec_cols].values.astype(np.float64)
+        src_wl = np.array([float(c) for c in spec_cols], dtype=np.float64)
+        dst_wl = np.array(train_wl, dtype=np.float64)
+        resampled = np.zeros((spec_data.shape[0], len(dst_wl)), dtype=np.float64)
+        for i in range(spec_data.shape[0]):
+            resampled[i] = np.interp(dst_wl, src_wl, spec_data[i],
+                                     left=np.nan, right=np.nan)
+        resampled = np.nan_to_num(resampled, nan=0.0)
+
+        result = pd.DataFrame(resampled,
+                              columns=[f'{wl:.6f}' for wl in train_wl])
+        print(f"[DualStream_MNF] 纯光谱重采样: {len(spec_cols)} → {len(train_wl)} 列")
+        print(f"[DualStream_MNF] pipeline.predict() 将自动完成 Physical+MNF 变换")
+        print(f"[特征对齐] 最终输入维度: {result.shape}")
+        return result.values
+
     def preprocess_spectra(self, spectra: pd.DataFrame) -> np.ndarray:
         """
         对光谱数据进行预处理 + 跨传感器光谱重采样。
@@ -477,34 +518,10 @@ class WaterQualityInference:
         metadata = self.loaded_model_data.get('metadata', {})
 
         # ═══════════════════════════════════════════════════════════
-        # ★ DualStream_MNF：纯光谱 → 重采样到训练波长 → pipeline.predict()
-        #   Pipeline 内置的 FeatureUnion[PhysicalExtractor + MNF]
-        #   自动计算物理指数和降维，无需外部补 WQI
+        # ★ DualStream_MNF：纯光谱重采样 → pipeline.predict() 全自动
         # ═══════════════════════════════════════════════════════════
         if actual_preprocess_method == "DualStream_MNF" and isinstance(model, Pipeline):
-            print("[DualStream_MNF] 推理：纯光谱重采样 → Pipeline 全自动处理")
-            train_wl = metadata.get('train_wavelengths', None)
-            if train_wl is not None and len(train_wl) > 0:
-                spec_cols = []
-                for c in spectra.columns:
-                    try:
-                        float(str(c)); spec_cols.append(c)
-                    except (ValueError, TypeError):
-                        pass
-                spec_data = spectra[spec_cols].values.astype(np.float64)
-                src_wl = np.array([float(c) for c in spec_cols], dtype=np.float64)
-                dst_wl = np.array(train_wl, dtype=np.float64)
-                resampled = np.zeros((spec_data.shape[0], len(dst_wl)), dtype=np.float64)
-                for i in range(spec_data.shape[0]):
-                    resampled[i] = np.interp(dst_wl, src_wl, spec_data[i],
-                                             left=np.nan, right=np.nan)
-                resampled = np.nan_to_num(resampled, nan=0.0)
-                spectra = pd.DataFrame(resampled,
-                    columns=[f'{wl:.6f}' for wl in train_wl])
-            print(f"[DualStream_MNF] 纯光谱重采样完成: {spectra.shape[1]} 列 "
-                  f"→ pipeline.predict() 将自动完成 Physical+MNF 变换")
-            print(f"[特征对齐] 最终输入维度: {spectra.shape}")
-            return spectra.values
+            return self._preprocess_dual_stream(spectra, metadata)
 
         train_wavelengths = metadata.get('train_wavelengths', None)
         # 旧模型无 train_wavelengths → 不做波长匹配，走下方分支 B 的 linspace 路径