diff --git a/src/gui/panels/step12_viz_panel.py b/src/gui/panels/step12_viz_panel.py
index 606eb85..14fa2c3 100644
--- a/src/gui/panels/step12_viz_panel.py
+++ b/src/gui/panels/step12_viz_panel.py
@@ -46,16 +46,31 @@ def _viz_training_spectra_csv_path(work_path: Path) -> Path:
 
 def _viz_infer_wavelength_start_column(df: pd.DataFrame) -> Union[str, int]:
     """推断光谱起始列（training_spectra 通常以波长数值为列名，未必含 UTM_Y）。"""
+    import re
+    # 兼容前缀：nm / nm_ / wavelength / wl / λ / wave_len（大小写不敏感）
+    _PREFIX_RE = re.compile(r"^(?:nm|wavelength|wl|\u03bb|wave_len)[_\s\-]*", re.IGNORECASE)
     for i, col in enumerate(df.columns):
         name = str(col).strip().lstrip("\ufeff")
+        # 1) 直接纯数字（如 "700.0"）
         try:
             v = float(name)
         except ValueError:
-            continue
-        if 200.0 <= v <= 3000.0:
+            v = None
+        if v is not None and 200.0 <= v <= 3000.0:
             return i
+        # 2) 剥前缀后再试（如 "nm_700"、"wavelength_800.0"、"WL 900"）
+        stripped = _PREFIX_RE.sub("", name)
+        if stripped != name:
+            try:
+                v = float(stripped)
+            except ValueError:
+                continue
+            if 200.0 <= v <= 3000.0:
+                return i
     if "UTM_Y" in df.columns:
         return "UTM_Y"
+    # 兜底分支触发时打印所有列名，便于下次失败时一眼看到根因
+    print(f"DEBUG: 尝试解析的列名: {df.columns.tolist()}")
     return 0