Python Opencv Detecting Circles from the Binary Image by Using Hough Circle Transform

Question

I am trying trying to split video which is threshold applied into frames then trying to find the circles in the image. But I searched that converting binary image to grayscale is not possible. I searched about Hough circle and the method only can take grayscale images. Hough Lines can work on binary images but hough circles could not. Is there any advice to use thresholded images in hough circle method? Please help me.

ps: I am adding the code and image that purpose is to find the circles in the thresholded image.

while videoCapture.isOpened(): #Begins to detect the captures in video by frames

    ret, image = videoCapture.read()
    print("image capture opened")

    if ret == True:

        #rgb = cv2.cvtColor(image, cv2.COLOR_HLS2RGB)
        gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
        bgr = cv2.cvtColor(gray, cv2.COLOR_GRAY2BGR) #converting video color to gray

        print("gray scaled image\n")
        frameCounter = frameCounter + 1
        circles = cv2.HoughCircles(gray, cv2.HOUGH_GRADIENT, 1, 20, param1=71, param2=70, minRadius=0, maxRadius=0)
        if circles is not None:

            print("Hough Circle on each frame")
            circles = np.uint16(np.around(circles))
            for i in circles[0, :]:
                cv2.circle(bgr, (i[0],i[1]), i[2], (0, 255, 0), 2) #Outer circle in the image
                cv2.circle(bgr, (i[0],i[1]), 2, (0, 0, 255), 3) #inner circle center
                print("inner outer circle draw")

            cv2.imwrite(outputDir + "/%d.jpg" % (frameCounter), bgr) #Saving frame to the output directory
        else :
            print('Circle could not find')
            cv2.imwrite(outputDir + "/%d.jpg" % (frameCounter), bgr)  # Saving frame to the output directory

        print("image saved to directory")

        videoOutput.write(bgr)

        if(frameCounter > (frameLength-1)):
            endTime = time.time()
            videoCapture.release()
            videoOutput.release()
            print("Converting video took %d seconds." % (endTime-startTime))
            break
    else:
       break

---

Answer 1

Try using convertTo rather than cvtColor. This example works:

cv::Mat image = imread("binary.bmp");
cv::Mat outImage;
image.convertTo(outImage, CV_8U);
imwrite("grayscale.bmp", outImage);

P.S.: you'll still have to play with param1 and param2 parameters of HoughCircles depending on how "circular" are the circles you want to detect, in your case it's really not a perfect circle. It would be much easyer to start detecting coins in an image to practice.