Real time circle detection using OpenCL accelerated Hough transform
This video is a demonstration of CMSoft's GPU accelerated Hough transform algorithm to be presented as an OpenCLTemplate case study.
Source code can be found here:
http://cmsoft.com.br/download/CLHough...
To use it: select a picture file containing the geometry that you would like to find. Click Load image - median filter - Sobel operator - thinning - Generate geometry - webcam.
Generate geometries from images of your camera for better results.
In our tests we could detect 37 coins in the acceptable rate of 7 fps (real time circle detection) using the Hough transform. Notice the robustness of the detection.
The image taken from the webcam is preprocessed with a Median filter followed by Sobel edge detection and a thinning algorithm. Then, a Generalized Hough Transform is performed using a custom, user-defined shape.
Soundtrack is a small cut from Trenzinho Caipira.
Visit http://www.cmsoft.com.br for more details.
Please check
http://www.cmsoft.com.br/index.php?op...
if you want to test our implementation.