fix: 提升目录OCR识别质量 - DPI 600 + 二值化增强

- OCR_DPI 150→600，高DPI渲染后缩放到800px保留文字细节
- _compress_image 增加二值化（灰度→阈值220→PNG），浅灰文字变黑加粗
- 同步 core/ 和 utils_test/ 两处

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

core/construction_review/component/minimal_pipeline/toc_detector.py

@@ -65,8 +65,8 @@ class TOCCatalogExtractor:
     MAX_CHECK_PAGES = 50
     DPI = 150
 
-    # OCR 配置（目录页使用更低DPI避免请求过大）
-    OCR_DPI = 150
+    # OCR 配置（高 DPI 渲染后缩放到 800px，确保目录文字清晰）
+    OCR_DPI = 600
     MAX_SHORT_EDGE = 800
     JPEG_QUALITY = 85
     MAX_IMAGE_SIZE_MB = 5
@@ -371,9 +371,13 @@ class TOCCatalogExtractor:
                 new_size = (int(img.width * ratio), int(img.height * ratio))
                 img = img.resize(new_size, Image.Resampling.LANCZOS)
 
+            # 二值化增强：将浅灰文字变黑，提高 OCR 识别率
+            img = img.convert('L')
+            img = img.point(lambda x: 0 if x < 220 else 255)
+            img = img.convert('RGB')
+
             buffer = io.BytesIO()
-            quality = self.JPEG_QUALITY if not force_smaller else 75
-            img.save(buffer, format='JPEG', quality=quality, optimize=True)
+            img.save(buffer, format='PNG', optimize=True)
             return buffer.getvalue()
 
         except Exception as e:

utils_test/minimal_pipeline/_toc_detector.py

@@ -65,8 +65,8 @@ class TOCCatalogExtractor:
     MAX_CHECK_PAGES = 50
     DPI = 150
 
-    # OCR 配置（目录页使用更低DPI避免请求过大）
-    OCR_DPI = 150
+    # OCR 配置（高 DPI 渲染后缩放到 800px，确保目录文字清晰）
+    OCR_DPI = 600
     MAX_SHORT_EDGE = 800
     JPEG_QUALITY = 85
     MAX_IMAGE_SIZE_MB = 5
@@ -371,9 +371,13 @@ class TOCCatalogExtractor:
                 new_size = (int(img.width * ratio), int(img.height * ratio))
                 img = img.resize(new_size, Image.Resampling.LANCZOS)
 
+            # 二值化增强：将浅灰文字变黑，提高 OCR 识别率
+            img = img.convert('L')
+            img = img.point(lambda x: 0 if x < 220 else 255)
+            img = img.convert('RGB')
+
             buffer = io.BytesIO()
-            quality = self.JPEG_QUALITY if not force_smaller else 75
-            img.save(buffer, format='JPEG', quality=quality, optimize=True)
+            img.save(buffer, format='PNG', optimize=True)
             return buffer.getvalue()
 
         except Exception as e: