sub_ocr.py 5.8 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187
  1. # This Python file uses the following encoding: utf-8
  2. # @author runhey
  3. # github https://github.com/runhey
  4. import cv2
  5. import re
  6. from datetime import timedelta
  7. from module.ocr.ppocr import TextSystem
  8. from module.exception import ScriptError
  9. from module.base.utils import area_pad, crop, float2str
  10. from module.ocr.base_ocr import BaseCor, OcrMode, OcrMethod
  11. from module.ocr.utils import merge_area
  12. from module.logger import logger
  13. class Full(BaseCor):
  14. """
  15. 这个类适用于大ROI范围的文本识别。可以支持多条文本识别, 默认不支持竖方向的文本识别
  16. """
  17. def after_process(self, result):
  18. return result
  19. def ocr_full(self, image, keyword: str=None) -> tuple:
  20. """
  21. 检测整个图片的文本,并对结果进行过滤。返回的是匹配到的keyword的左边。如果没有匹配到返回(0, 0, 0, 0)
  22. :param image:
  23. :param keyword:
  24. :return:
  25. """
  26. boxed_results = self.detect_and_ocr(image)
  27. if not boxed_results:
  28. return 0, 0, 0, 0
  29. index_list = self.filter(boxed_results, keyword)
  30. # 如果一个都没有匹配到
  31. if not index_list:
  32. return 0, 0, 0, 0
  33. # 如果匹配到了多个,则合并所有的坐标,返回合并后的坐标
  34. if len(index_list) > 1:
  35. area_list = [(
  36. boxed_results[index].box[0, 0], # x
  37. boxed_results[index].box[0, 1], # y
  38. boxed_results[index].box[1, 0] - boxed_results[index].box[0, 0], # width
  39. boxed_results[index].box[2, 1] - boxed_results[index].box[0, 1], # height
  40. ) for index in index_list]
  41. box = merge_area(area_list)
  42. return box
  43. else:
  44. box = boxed_results[index_list[0]].box
  45. return box[0, 0], box[0, 1], box[1, 0] - box[0, 0], box[2, 1] - box[0, 1]
  46. class Single(BaseCor):
  47. """
  48. 这个类使用于单行文本识别(所识别的ROI不会动)
  49. """
  50. def after_process(self, result):
  51. return result
  52. def ocr_single(self, image) -> str:
  53. """
  54. 检测某个固定位置的roi的文本。可以是横方向也可以是竖方向
  55. :param image:
  56. :return:
  57. """
  58. if self.roi:
  59. result = self.ocr_single_line(image)
  60. if result != "":
  61. return result
  62. # 如果没有识别到,就尝试逆时针旋转90度再识别一次。这个时候考虑到可能是竖方向的文本
  63. image = cv2.rotate(image, cv2.ROTATE_90_COUNTERCLOCKWISE)
  64. result = self.ocr_single_line(image)
  65. if result != "":
  66. return result
  67. # 如果还是没有识别到。那可能就是真的没有识别到了
  68. return ""
  69. else:
  70. raise ScriptError("Roi is empty")
  71. class Digit(Single):
  72. def after_process(self, result):
  73. print('this is digit after_process')
  74. result = super().after_process(result)
  75. result = result.replace('I', '1').replace('D', '0').replace('S', '5')
  76. result = result.replace('B', '8')
  77. prev = result
  78. result = int(result) if result else 0
  79. if str(result) != prev:
  80. logger.warning(f'OCR {self.name}: Result "{prev}" is revised to "{result}"')
  81. return result
  82. def ocr_digit(self, image) -> int:
  83. """
  84. 返回数字
  85. :param image:
  86. :return:
  87. """
  88. result = self.ocr_single(image)
  89. if result == "":
  90. return 0
  91. else:
  92. return int(result)
  93. class DigitCounter(Single):
  94. def after_process(self, result):
  95. result = super().after_process(result)
  96. result = result.replace('I', '1').replace('D', '0').replace('S', '5')
  97. result = result.replace('B', '8')
  98. return result
  99. @classmethod
  100. def ocr_str_digit_counter(cls, result: str) -> tuple[int, int, int]:
  101. result = re.search(r'(\d+)/(\d+)', result)
  102. if result:
  103. result = [int(s) for s in result.groups()]
  104. current, total = int(result[0]), int(result[1])
  105. current = min(current, total)
  106. return current, total - current, total
  107. else:
  108. logger.warning(f'Unexpected ocr result: {result}')
  109. return 0, 0, 0
  110. def ocr_digit_counter(self, image) -> tuple[int, int, int]:
  111. """
  112. 获取计数的结果
  113. :param image:
  114. :return: 例如 14/15,返回 (14, 1, 15) 。如果没有识别到,返回 (0, 0, 0)
  115. """
  116. result = self.ocr_single(image)
  117. if result == "":
  118. return 0, 0, 0
  119. return self.ocr_str_digit_counter(result)
  120. class Duration(Single):
  121. def after_process(self, result):
  122. result = super().after_process(result)
  123. result = result.replace('I', '1').replace('D', '0').replace('S', '5')
  124. result = result.replace('B', '8')
  125. return result
  126. @staticmethod
  127. def parse_time(string):
  128. """
  129. Args:
  130. string (str): `01:30:00`
  131. Returns:
  132. datetime.timedelta:
  133. """
  134. result = re.search(r'(\d{1,2}):?(\d{2}):?(\d{2})', string)
  135. if result:
  136. result = [int(s) for s in result.groups()]
  137. return timedelta(hours=result[0], minutes=result[1], seconds=result[2])
  138. else:
  139. logger.warning(f'Invalid duration: {string}')
  140. return timedelta(hours=0, minutes=0, seconds=0)
  141. def ocr_duration(self, image) -> timedelta:
  142. """
  143. :param image:
  144. :return:
  145. """
  146. result = self.ocr_single(image)
  147. if result == "":
  148. return timedelta(hours=0, minutes=0, seconds=0)
  149. return self.parse_time(result)
  150. if __name__ == '__main__':
  151. d = DigitCounter.ocr_str_digit_counter('14/77')
  152. print(d)
  153. u = Duration.parse_time('01:30:00')
  154. print(u)