|
|
@@ -378,6 +378,19 @@ class MinerUWiredTableRecognizer:
|
|
|
|
|
|
# Step 2: 使用连通域法提取单元格 (替换了原来的投影法)
|
|
|
debug_prefix = f"{dbg.prefix}_grid" if dbg.prefix else "grid"
|
|
|
+
|
|
|
+ # 计算 OCR 文本容差：取最小行高的 50%，无有效 OCR 时回退为 0
|
|
|
+ ocr_heights = []
|
|
|
+ for ocr in ocr_boxes or []:
|
|
|
+ bbox = ocr.get("bbox", [])
|
|
|
+ if len(bbox) >= 4:
|
|
|
+ height = bbox[3] - bbox[1]
|
|
|
+ if height > 0:
|
|
|
+ ocr_heights.append(height)
|
|
|
+ if ocr_heights:
|
|
|
+ ocr_text_pixel_tolerance = min(ocr_heights) * 0.5
|
|
|
+ else:
|
|
|
+ ocr_text_pixel_tolerance = 10.0
|
|
|
|
|
|
# 传入原图的实际尺寸和裁剪padding
|
|
|
bboxes = self.grid_recovery.compute_cells_from_lines(
|
|
|
@@ -402,6 +415,7 @@ class MinerUWiredTableRecognizer:
|
|
|
table_image=table_image,
|
|
|
unet_cells=bboxes,
|
|
|
ocr_boxes=ocr_boxes or [],
|
|
|
+ ocr_text_pixel_tolerance=ocr_text_pixel_tolerance,
|
|
|
pdf_type=pdf_type,
|
|
|
debug_dir=debug_dir,
|
|
|
debug_prefix=debug_prefix
|
|
|
@@ -425,7 +439,10 @@ class MinerUWiredTableRecognizer:
|
|
|
|
|
|
# Step 3: 重建网格结构 (计算 row, col, rowspan, colspan)
|
|
|
# OCR补偿已在Step 2中完成，这里仅做网格重建
|
|
|
- merged_cells = self.grid_recovery.recover_grid_structure(bboxes)
|
|
|
+ merged_cells = self.grid_recovery.recover_grid_structure(
|
|
|
+ bboxes,
|
|
|
+ ocr_text_pixel_tolerance=ocr_text_pixel_tolerance
|
|
|
+ )
|
|
|
|
|
|
# Step 3.5: 可视化逻辑结构 (新增)
|
|
|
if self.debug_utils.debug_is_on("save_grid_structure", dbg):
|
