feat: 更新数据处理逻辑以使用PaddleOCR_VL的旋转角度和原始图像尺寸

merger/data_processor.py

@@ -225,17 +225,13 @@ class DataProcessor:
         Returns:
             🎯 MinerU 格式的合并数据（统一输出格式）
         """
-        merged_data = []
-        paddle_pointer = 0
-        last_matched_index = 0
-        
         # 🎯 获取旋转角度和原始图像尺寸
-        rotation_angle = self._get_rotation_angle_from_vl(paddleocr_vl_data)
-        vl_orig_image_size = None
+        vl_rotation_angle = self._get_rotation_angle_from_vl(paddleocr_vl_data)
+        vl_orig_image_size = (0,0)
         
-        if rotation_angle != 0:
+        if vl_rotation_angle != 0:
             vl_orig_image_size = self._get_original_image_size_from_vl(paddleocr_vl_data)
-            print(f"🔄 PaddleOCR_VL 检测到旋转角度: {rotation_angle}°")
+            print(f"🔄 PaddleOCR_VL 检测到旋转角度: {vl_rotation_angle}°")
             print(f"📐 原始图像尺寸: {vl_orig_image_size[0]} x {vl_orig_image_size[1]}")
         
         # 提取 parsing_res_list
@@ -250,8 +246,8 @@ class DataProcessor:
     
         for item in parsing_res_list:
             # 🎯 先转换 bbox 坐标（如果需要）
-            if rotation_angle != 0 and orig_image_size:
-                item = self._transform_vl_block_bbox(item, rotation_angle, orig_image_size)
+            if vl_rotation_angle != 0 and orig_image_size:
+                item = self._transform_vl_block_bbox(item, vl_rotation_angle, vl_orig_image_size)
             converted_item = self._convert_paddleocr_vl_to_mineru(item)
             if converted_item:
                 mineru_format_data.append(converted_item)