Update utils/ocr_utils.py

utils/ocr_utils.py

@@ -5,54 +5,36 @@ from PIL import Image, ImageDraw, ImageFont
 ocr_model = PaddleOCR(use_textline_orientation=True, lang='ch')
 
 def group_nearby_boxes(lines, max_y_gap=50):
-    """
-    Groups OCR boxes if they are vertically close.
-    Returns: list of dicts with 'polygons' and 'texts'
-    """
-    boxes = []
-    for line in lines:
-        poly = line[0]
-        text = line[1][0]
-
-        if not text.strip():
+    groups = []
+    used = set()
+
+    def is_valid_polygon(poly):
+        return isinstance(poly, (list, tuple)) and all(
+            isinstance(p, (list, tuple)) and len(p) == 2 for p in poly
+        )
+
+    for i, (poly_i, text_i) in enumerate(lines):
+        if i in used or not is_valid_polygon(poly_i):
             continue
 
-        xs, ys = zip(*poly)
-        x_min, x_max = min(xs), max(xs)
-        y_min, y_max = min(ys), max(ys)
+        group = [(poly_i, text_i)]
+        used.add(i)
 
-        boxes.append({
-            "polygon": poly,
-            "text": text,
-            "y_center": (y_min + y_max) / 2,
-            "y_min": y_min,
-            "y_max": y_max,
-        })
+        xi_min, yi_min = min(pt[1] for pt in poly_i), min(pt[0] for pt in poly_i)
+        xi_max, yi_max = max(pt[1] for pt in poly_i), max(pt[0] for pt in poly_i)
 
-    # Sort by vertical position (top to bottom)
-    boxes.sort(key=lambda b: b["y_center"])
+        for j, (poly_j, text_j) in enumerate(lines):
+            if j in used or not is_valid_polygon(poly_j):
+                continue
 
-    groups = []
-    current_group = [boxes[0]]
-
-    for curr in boxes[1:]:
-        last = current_group[-1]
-        vertical_gap = curr["y_min"] - last["y_max"]
-
-        if vertical_gap < max_y_gap:
-            current_group.append(curr)
-        else:
-            groups.append({
-                "polygons": [b["polygon"] for b in current_group],
-                "texts": [b["text"] for b in current_group],
-            })
-            current_group = [curr]
-
-    if current_group:
-        groups.append({
-            "polygons": [b["polygon"] for b in current_group],
-            "texts": [b["text"] for b in current_group],
-        })
+            xj_min, yj_min = min(pt[1] for pt in poly_j), min(pt[0] for pt in poly_j)
+            xj_max, yj_max = max(pt[1] for pt in poly_j), max(pt[0] for pt in poly_j)
+
+            if abs(yj_min - yi_min) < max_y_gap or abs(yj_max - yi_max) < max_y_gap:
+                group.append((poly_j, text_j))
+                used.add(j)
+
+        groups.append(group)
 
     return groups