Clutter Problem with Hough Transform for line detection

asked 2013-01-25 02:45:30 -0500

fxhtc gravatar image

I am working on a computer vision project. One step of the project requires me to detect the strips inside an image. I've tried the combination of Canny + Hough transform. The problem is that I usually get several lines detected as the boundary of the strips, which causes problem for me. It seems to be a known problem for hough transform, but is there any method I can use to eliminate the repetitive detection of the same line? I am using the kind of the euclidean distance between two lines (given rho and theta) to determine if they are actually detecting the same line, but this method does not work very well.

edit retag flag offensive close merge delete

Comments

Please put your sample image here.

Mostafa Sataki gravatar imageMostafa Sataki ( 2013-01-25 05:19:01 -0500 )edit

Have you tried using the probabilistic Hough transform, too? Supposingly It gives less but more relieble lines. I would also experiment with the canny thresholds, actually it should give only one line, but maybe you used wrong values.

Guanta gravatar imageGuanta ( 2013-01-25 07:19:56 -0500 )edit