我有一组图像,其中的框和圆圈包含数字。一些盒子和圆圈没有在所有边上连接,线条上有一些间隙。我想填补这个空白(如图所示),任何人都知道使用 OpenCV ......?我的代码在这里………… self.res = cv2.imread("1.jpg") self.store_path = "../storage-data/" #define kernal value kernel = np.ones((2,2),np.uint8) #grayscale gray = cv2.cvtColor(self.image, cv2.COLOR_BGR2GRAY) dilate = cv2.dilate(gray,kernel,iterations = 1) dilate1 = cv2.dilate(dilate,kernel,iterations = 1) dilate1 = cv2.morphologyEx(dilate1, cv2.MORPH_OPEN, kernel) #Canny canny = cv2.Canny(dilate1,160,160,3) dilate = cv2.dilate(canny,kernel,iterations = 1) #Gaussian Blurring blur = cv2.GaussianBlur(dilate,(5,5),0) erode = cv2.dilate(blur,kernel,iterations = 1) blur = cv2.GaussianBlur(erode,(5,5),1) blur = cv2.morphologyEx(blur, cv2.MORPH_CLOSE, kernel) ret, thresh = cv2.threshold(blur,127,255,cv2.THRESH_BINARY) blur = cv2.GaussianBlur(thresh,(5,5),1) ret1, thresh1 = cv2.threshold(blur,127,255,cv2.THRESH_BINARY) opening = cv2.morphologyEx(thresh1, cv2.MORPH_OPEN, kernel) cv2.imwrite(self.store_path + 'opening.jpg', opening) contours, hierarchy = cv2.findContours(opening,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)
2 回答

子衿沉夜
TA贡献1828条经验 获得超3个赞
一次膨胀就足够了,关键是使用垂直内核,尝试更改内核的尺寸,例如 kernel = np.ones((5,2),np.uint8) 检查 1 次膨胀后的结果,它应该缩小差距.

繁华开满天机
TA贡献1816条经验 获得超4个赞
您可以尝试使用轮廓吗?轮廓可以简单地解释为连接所有连续点(沿边界)的曲线,具有相同的颜色或强度。首先找到所有的计数,然后将轮廓绘制到原始图像上。这可能会解决问题
contours, _ = cv2.findContours(canny, 2,2)
length = len(contours)
for i in range(len(contours)):
cv2.drawContours(img, contours, i, (0,255,0), 2, cv2.LINE_8)
添加回答
举报
0/150
提交
取消