Переглянути джерело

feat(module): Recognize special characters and convert them to numbers

huangrunheng 1 рік тому
батько
коміт
13aa93b19a
5 змінених файлів з 66 додано та 3 видалено
  1. 18 3
      module/atom/ocr.py
  2. 1 0
      module/ocr/base_ocr.py
  3. 44 0
      module/ocr/sub_ocr.py
  4. 1 0
      requirements-in.txt
  5. 2 0
      requirements.txt

+ 18 - 3
module/atom/ocr.py

@@ -6,12 +6,12 @@ import numpy as np
 import cv2
 import cv2
 
 
 from module.ocr.base_ocr import BaseCor, OcrMode, OcrMethod
 from module.ocr.base_ocr import BaseCor, OcrMode, OcrMethod
-from module.ocr.sub_ocr import Full, Single, Digit, DigitCounter, Duration
+from module.ocr.sub_ocr import Full, Single, Digit, DigitCounter, Duration, Quantity
 from module.logger import logger
 from module.logger import logger
 
 
 
 
 
 
-class RuleOcr(Digit, DigitCounter, Duration, Single, Full):
+class RuleOcr(Digit, DigitCounter, Duration, Single, Full, Quantity):
 
 
     def __init__(self, *args, **kwargs):
     def __init__(self, *args, **kwargs):
         super().__init__(*args, **kwargs)
         super().__init__(*args, **kwargs)
@@ -24,6 +24,7 @@ class RuleOcr(Digit, DigitCounter, Duration, Single, Full):
             case OcrMode.DIGIT: return Digit.after_process(self, result)
             case OcrMode.DIGIT: return Digit.after_process(self, result)
             case OcrMode.DIGITCOUNTER: return DigitCounter.after_process(self, result)
             case OcrMode.DIGITCOUNTER: return DigitCounter.after_process(self, result)
             case OcrMode.DURATION: return Duration.after_process(self, result)
             case OcrMode.DURATION: return Duration.after_process(self, result)
+            case OcrMode.QUANTITY: return Quantity.after_process(self, result)
             case _: return result
             case _: return result
 
 
     def ocr(self, image, keyword=None):
     def ocr(self, image, keyword=None):
@@ -34,6 +35,7 @@ class RuleOcr(Digit, DigitCounter, Duration, Single, Full):
             case OcrMode.DIGIT: return Digit.ocr_digit(self, image)
             case OcrMode.DIGIT: return Digit.ocr_digit(self, image)
             case OcrMode.DIGITCOUNTER: return DigitCounter.ocr_digit_counter(self, image)
             case OcrMode.DIGITCOUNTER: return DigitCounter.ocr_digit_counter(self, image)
             case OcrMode.DURATION: return Duration.ocr_duration(self, image)
             case OcrMode.DURATION: return Duration.ocr_duration(self, image)
+            case OcrMode.QUANTITY: return Quantity.ocr_quantity(self, image)
             case _: return None
             case _: return None
 
 
     def coord(self) -> tuple:
     def coord(self) -> tuple:
@@ -55,5 +57,18 @@ class RuleOcr(Digit, DigitCounter, Duration, Single, Full):
 
 
 
 
 if __name__ == "__main__":
 if __name__ == "__main__":
-    pass
+    O_MALL_RESOURCE_1 = RuleOcr(roi=(144, 7, 100, 43), area=(144, 7, 100, 43), mode="Quantity", method="Default",
+                                keyword="", name="mall_resource_1")
+    O_MALL_RESOURCE_2 = RuleOcr(roi=(326, 8, 124, 39), area=(326, 8, 124, 39), mode="Quantity", method="Default",
+                                keyword="", name="mall_resource_2")
+    O_MALL_RESOURCE_3 = RuleOcr(roi=(533, 9, 107, 38), area=(533, 9, 107, 38), mode="Quantity", method="Default",
+                                keyword="", name="mall_resource_3")
+    O_MALL_RESOURCE_4 = RuleOcr(roi=(739, 8, 100, 39), area=(739, 8, 100, 39), mode="Quantity", method="Default",
+                                keyword="", name="mall_resource_4")
+    O_MALL_RESOURCE_5 = RuleOcr(roi=(935, 11, 100, 37), area=(935, 11, 100, 37), mode="Quantity", method="Default",
+                                keyword="", name="mall_resource_5")
+    O_MALL_RESOURCE_6 = RuleOcr(roi=(1129, 6, 100, 41), area=(1129, 6, 100, 41), mode="Quantity", method="Default",
+                                keyword="", name="mall_resource_6")
+    image = cv2.imread(r"E:\2025-01-16225353.png")
+    print(O_MALL_RESOURCE_5.ocr_quantity(image))
 
 

+ 1 - 0
module/ocr/base_ocr.py

@@ -39,6 +39,7 @@ class OcrMode(Enum):
     DIGIT = 3  # str: "Digit"
     DIGIT = 3  # str: "Digit"
     DIGITCOUNTER = 4  # str: "DigitCounter"
     DIGITCOUNTER = 4  # str: "DigitCounter"
     DURATION = 5  # str: "Duration"
     DURATION = 5  # str: "Duration"
+    QUANTITY = 6  # str: "Quantity"
 
 
 class OcrMethod(Enum):
 class OcrMethod(Enum):
     DEFAULT = 1  # str: "Default"
     DEFAULT = 1  # str: "Default"

+ 44 - 0
module/ocr/sub_ocr.py

@@ -4,6 +4,7 @@
 
 
 import cv2
 import cv2
 import re
 import re
+import cn2an
 
 
 from datetime import timedelta
 from datetime import timedelta
 
 
@@ -196,6 +197,49 @@ class Duration(Single):
 
 
         return self.parse_time(result)
         return self.parse_time(result)
 
 
+class Quantity(BaseCor):
+    """
+    专门用于识别超级多的数量,不支持多个区域的识别
+    可支持负数
+    比如:”6.33亿“ ”1.2万“ “53万/100” -> 530,000
+    """
+    def after_process(self, result):
+        result = super().after_process(result)
+        result = result.replace('I', '1').replace('D', '0').replace('S', '5')
+        result = result.replace('B', '8').replace('?', '2').replace('?', '2').replace('d', '6')
+        result = [char
+                  for char in result
+                  if char.isdigit() or char == '.' or char == '/' or char == '万' or char == '亿' or char == '千']
+        result = ''.join(result)
+
+        if '/' in result:
+            result_split = result.split('/')
+            result = result_split[0]
+        result = cn2an.cn2an(result, 'smart')
+
+        try:
+            result = int(result)
+        except ValueError:
+            logger.warning(f'[{self.name}]: Invalid quantity: {result}')
+            result = 0
+        return result
+
+    def ocr_quantity(self, image) -> int:
+        """
+        返回数量
+        :param image:
+        :return:
+        """
+        boxed_results = self.detect_and_ocr(image)
+        if not boxed_results:
+            logger.warning(f'[{self.name}]: No text detected')
+            return 0
+
+        box = boxed_results[0].box
+        self.area = box[0, 0] + self.roi[0], box[0, 1] + self.roi[1], box[1, 0] - box[0, 0], box[2, 1] - box[0, 1]
+        return boxed_results[0].ocr_text
+
+
 
 
 if __name__ == '__main__':
 if __name__ == '__main__':
     import cv2
     import cv2

+ 1 - 0
requirements-in.txt

@@ -12,6 +12,7 @@ pywin32==306
 
 
 # OCR
 # OCR
 ppocr-onnx==0.0.3.9
 ppocr-onnx==0.0.3.9
+cn2an==0.5.23
 
 
 # Web
 # Web
 paho-mqtt==1.6.1
 paho-mqtt==1.6.1

+ 2 - 0
requirements.txt

@@ -17,6 +17,7 @@ cffi==1.16.0              # via cryptography, gevent
 charset-normalizer==3.3.2  # via requests
 charset-normalizer==3.3.2  # via requests
 cigam==0.0.3              # via apkutils2
 cigam==0.0.3              # via apkutils2
 click==8.1.7              # via uvicorn
 click==8.1.7              # via uvicorn
+cn2an==0.5.23             # via -r requirements-in.txt
 colorama==0.4.6           # via click, logzero, tqdm
 colorama==0.4.6           # via click, logzero, tqdm
 coloredlogs==15.0.1       # via onnxruntime
 coloredlogs==15.0.1       # via onnxruntime
 cryptography==42.0.8      # via -r requirements-in.txt
 cryptography==42.0.8      # via -r requirements-in.txt
@@ -49,6 +50,7 @@ packaging==20.9           # via deprecation, onnxruntime, uiautomator2
 paho-mqtt==1.6.1          # via -r requirements-in.txt
 paho-mqtt==1.6.1          # via -r requirements-in.txt
 pillow==10.2.0            # via ppocr-onnx, uiautomator2
 pillow==10.2.0            # via ppocr-onnx, uiautomator2
 ppocr-onnx==0.0.3.9       # via -r requirements-in.txt
 ppocr-onnx==0.0.3.9       # via -r requirements-in.txt
+proces==0.1.7             # via cn2an
 progress==1.6             # via uiautomator2
 progress==1.6             # via uiautomator2
 protobuf==4.25.1          # via onnxruntime
 protobuf==4.25.1          # via onnxruntime
 psutil==6.1.1             # via -r requirements-in.txt
 psutil==6.1.1             # via -r requirements-in.txt