添加形近字处理

photo_mask/__init__.py

@@ -21,5 +21,10 @@ NAME_KEYS = [
 ]
 ID_CARD_NUM_KEYS = [{"key": "身份证号", "length": 19}, ]
 
+# 形近字，只对姓名涂抹生效
+SIMILAR_CHAR = {
+    "凤": ["风"]
+}
+
 # 如果不希望识别出空格，可以设置use_space_char=False。做此项设置一定要测试，2.7.3版本此项设置有bug，会导致识别失败
 OCR = PaddleOCR(use_angle_cls=False, show_log=False, det_db_box_thresh=0.3)

photo_mask/photo_mask.py

@@ -7,7 +7,7 @@ from sqlalchemy import update, and_
 
 from db import MysqlSession
 from db.mysql import ZxPhrec, ZxPhhd
-from photo_mask import OCR, PHHD_BATCH_SIZE, SLEEP_MINUTES, NAME_KEYS, ID_CARD_NUM_KEYS
+from photo_mask import OCR, PHHD_BATCH_SIZE, SLEEP_MINUTES, NAME_KEYS, ID_CARD_NUM_KEYS, SIMILAR_CHAR
 from ucloud import BUCKET, ufile
 from util import image_util, util
 
@@ -103,10 +103,19 @@ def get_mask_layout(image, name, id_card_num):
             # 无识别结果
             return result
         else:
-            # 涂抹
+            # 处理形近字
+            name_list = [name]
+            # 移除名字中重复的字
+            unique_name = "".join(dict.fromkeys(name))
+            for char in unique_name:
+                if char in SIMILAR_CHAR:
+                    for sc in SIMILAR_CHAR[char]:
+                        name_list.append(name.replace(char, sc))
+
             for layout in layouts:
                 find_name_by_key = True
                 find_id_card_num_by_key = True
+                for name in name_list:
                     if name in layout[1]:
                         result += find_boxes(name, layout, improve=True, image_path=temp_file.name)
                         find_name_by_key = False
@@ -146,6 +155,7 @@ def mask_photo(img_url, name, id_card_num, color=(255, 255, 255)):
     is_masked = False
     # 打开图片
     image = image_util.read(img_url)
+    # image = image_util.rotate(image, 180)
     split_results = image_util.split(image)
     for split_result in split_results:
         to_mask_img, x_offset, y_offset = handle_image_for_mask(split_result)