fix: PDF坐标对齐+金额解析修复+x过滤器移除

core/parser.py

@@ -195,9 +195,9 @@ class K3VoucherParser:
                     result['period'] = int(m.group(1))
             elif '合计' in t and any(c in t for c in '零壹贰叁肆伍陆柒捌玖拾佰仟万亿'):
                 result['total_str'] = t
-                m = re.search(r'([\d,]+\.\d{2})', t.replace(',', ''))
+                m = re.search(r'(\d+\.\d{2})', re.sub(r'[^\d.]', '', t))
                 if m:
-                    result['total_amount'] = float(m.group(1).replace(',', ''))
+                    result['total_amount'] = float(m.group(1))
             elif t.startswith('制单') or t.startswith('制单人'):
                 result['preparer'] = t.split('：')[-1].split(':')[-1].strip()
             elif t.startswith('审核'):
@@ -233,8 +233,8 @@ class K3VoucherParser:
                 continue
             
             # x坐标：科目代码通常在表格中间区域
-            if cx < 1500 or cx > 2600:
-                continue
+            # 注：不设硬x过滤器，正则 \d{3,4}\.\d 和首位数 1000-9999 校验已足够精准
+            # 设 x 过滤器会误杀 PDF 版本（排版偏左）和少部分照片（x=1297 也有代码）
             
             t = text.strip()
             # 排除表头文字
@@ -415,11 +415,13 @@ class K3VoucherParser:
             if any(k in text for k in ['Manager', 'HONOR', 'Magic6', 'RMB']):
                 continue
             
-            m = re.search(r'([\d,]+\.\d{2})', text.replace(',', ''))
+            # 提取金额：去掉所有非数字字符（逗号/空格等），避免"11, 075.68"被解析为75.68
+            cleaned = re.sub(r'[^\d.]', '', text)
+            m = re.search(r'(\d+\.\d{2})', cleaned)
             if not m:
                 continue
-            
-            val = float(m.group(1).replace(',', ''))
+
+            val = float(m.group(1))
             # 排除年份
             if 1900 < val < 2100:
                 continue

kocr.py

@@ -220,6 +220,33 @@ def extract_pdf_pages(pdf_path, output_dir):
         new_name = f"_{pdf_name}_p{page_num}.jpg"
         new_path = os.path.join(output_dir, new_name)
         os.rename(old_path, new_path)
+        
+        # 统一图片尺寸：缩放+白边到 3072×4096（对齐照片坐标）
+        # 解析器硬编码了 x/y 坐标阈值（适配 3072×4096 的照片）
+        # PDF 图片尺寸和比例不同，必须统一到照片尺寸才不走样
+        try:
+            import cv2
+            img = cv2.imread(new_path)
+            if img is not None and (img.shape[1] != 3072 or img.shape[0] != 4096):
+                h, w = img.shape[:2]
+                # 计算缩放比例（保持宽高比，适配目标尺寸）
+                scale = min(3072 / w, 4096 / h)
+                new_w = int(w * scale)
+                new_h = int(h * scale)
+                # 缩放
+                resized = cv2.resize(img, (new_w, new_h), interpolation=cv2.INTER_CUBIC)
+                # 中心对齐，加白边补齐到 3072×4096
+                pad_top = (4096 - new_h) // 2
+                pad_bottom = 4096 - new_h - pad_top
+                pad_left = (3072 - new_w) // 2
+                pad_right = 3072 - new_w - pad_left
+                padded = cv2.copyMakeBorder(resized, pad_top, pad_bottom,
+                                            pad_left, pad_right,
+                                            cv2.BORDER_CONSTANT, value=[255, 255, 255])
+                cv2.imwrite(new_path, padded)
+        except Exception as e:
+            pass  # 缩放失败不影响主流程
+        
         page_label = f"第{page_num}页" if m else os.path.basename(pdf_path)
         result_list.append((new_path, page_label))