图片红色矩形框内文字识别

IS Tim

1. 使用的技术

   • OpenCV
   • PaddleOCR
   • Python 3

2. OpenCV
   安装依赖

   	pip install opencv-python
   

   图片红色矩形框识别

   def opencv_paddle():
       """
       opencv识别红色矩形框，paddle识别文字
       :return:
       """
       # 加载图像
       img = cv2.imread("file/17119755916493.png")
       # 转换为HSV色彩空间
       hsv_img = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
       # 定义红色的HSV范围
       lower_red = (0, 43, 46)
       upper_red = (10, 255, 255)  # 这个范围可能需要根据实际情况调整
       mask = cv2.inRange(hsv_img, lower_red, upper_red)
       # 使用形态学操作去除噪声，如膨胀与腐蚀
       kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
       mask = cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel)
       # 找到轮廓
       contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
       # 遍历轮廓，筛选出矩形轮廓
       rectangles = []
       for cnt in contours:
           x, y, w, h = cv2.boundingRect(cnt)
           if w > 50 and h > 30:  # 设定最小宽度和高度阈值
               rectangles.append((x, y, w, h))
       # 从原图像中裁剪出红色矩形区域
       cropped_images = []
       for rect in rectangles:
           x, y, w, h = rect
           cropped = img[y:y + h, x:x + w]
           cropped_images.append(cropped)
   

3. PaddleOCR
   安装依赖

   	pip install paddlepaddle 
   	pip install paddleocr  
   

   图片识别

       """
   paddle 图片文字识别
   :return:
   """
   ocr = PaddleOCR(use_angle_cls=True, use_gpu=False)  # 使用CPU预加载，不用GPU
   text = ocr.ocr("file/17119755916493.png", cls=True)  # 打开图片文件
   print(text)
   # 打印所有文本信息
   for t in text:
       print(t[1][0])
   

4. 完整实例

   from paddleocr import PaddleOCR
   import cv2
   from PIL import Image
   import paddleocr
   import numpy as np
   
   
   def opencv():
       """
       paddle 图片文字识别
       :return:
       """
       ocr = PaddleOCR(use_angle_cls=True, use_gpu=False)  # 使用CPU预加载，不用GPU
       text = ocr.ocr("file/17119755916493.png", cls=True)  # 打开图片文件
       print(text)
       # 打印所有文本信息
       for t in text:
           print(t[1][0])
   
   
   def opencv_paddle():
       """
       opencv识别红色矩形框，paddle识别文字
       :return:
       """
       # 加载图像
       img = cv2.imread("file/17119755916493.png")
       # 转换为HSV色彩空间
       hsv_img = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
       # 定义红色的HSV范围
       lower_red = (0, 43, 46)
       upper_red = (10, 255, 255)  # 这个范围可能需要根据实际情况调整
       mask = cv2.inRange(hsv_img, lower_red, upper_red)
       # 使用形态学操作去除噪声，如膨胀与腐蚀
       kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
       mask = cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel)
       # 找到轮廓
       contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
       # 遍历轮廓，筛选出矩形轮廓
       rectangles = []
       for cnt in contours:
           x, y, w, h = cv2.boundingRect(cnt)
           if w > 50 and h > 30:  # 设定最小宽度和高度阈值
               rectangles.append((x, y, w, h))
       # 从原图像中裁剪出红色矩形区域
       cropped_images = []
       for rect in rectangles:
           x, y, w, h = rect
           cropped = img[y:y + h, x:x + w]
           cropped_images.append(cropped)
   
       # 初始化PaddleOCR
       ocr = paddleocr.PaddleOCR(use_angle_cls=True, lang='ch')  # 如果是中文
   
       # 对每个裁剪出的矩形框进行文字识别
       for cropped in cropped_images:
           # 将OpenCV格式转换为PIL格式
           pil_image = Image.fromarray(cv2.cvtColor(cropped, cv2.COLOR_BGR2RGB))
           # 将PIL.Image转换为numpy数组
           np_image = np.array(pil_image)
           # 进行文字识别
           result = ocr.ocr(np_image)
   
           # 输出识别的文字
           for line in result:
               # 坐标，文字
               print(line)
   
       # 或者你可以保存每个裁剪区域并分别识别
       # for i, crop in enumerate(cropped_images):
       #     cv2.imwrite(f'crop_{i}.jpg', crop)
       #     # 然后针对每个保存的文件运行PaddleOCR识别
   
   
   if __name__ == '__main__':
       opencv_paddle()