0

I have been working on a problem to identify the moving ball in monochromatic images in various lightning conditions using opencv. Following are the steps I had done: 1. Blurred the image using a median/Gaussian blur 2. Thresholded the image using simple thresholding 3. Found out the edges in the image using Canny Edge detector 4. To the obtained image, used Hough circle transform to find all the circular shaped objects in the image and using the circle size thresholded the unwanted circles in the image.

It is giving good results in case of uniform lightning conditions, but in case of non-uniform lightning and in case of shadows, ball is not being detected properly. Can anyone suggest any proper method to detect the moving ball in monochromatic images.

Sai krishna
  • 168
  • 9
  • Have you tried an histogram normalization? – Matteo Ragni Jun 09 '18 at 12:47
  • No I havent tried it but I think it does not work because the ball is moving and lightning conditions are changing the ball is being covered by say net in some cases. – Sai krishna Jun 09 '18 at 12:54
  • Maybe I wan not clear enough: have you tried to do an histogram normalization before applying your algorithm to the frame? The histogram equalization is used to get rid of lighting issues. – Matteo Ragni Jun 09 '18 at 14:38
  • I had carried out the histogram equalization, not much improvement is observed. I had got an improvement in detection of the ball using morphologyEx with MORPH_TOPHAT parameter. Now the problem i'm facing is, shadows are being detected along with the ball. How can I overcome the detection of shadows in the above case. – Sai krishna Jun 19 '18 at 10:31

0 Answers0