Hi, I have discovered kmeans, and am using it like this:
Mat labels, centers;
int attempts = 50, clusterCount = 4, flags = cv::KMEANS_PP_CENTERS;
TermCriteria tc;
kmeans(centroids, clusterCount, labels, tc, attempts, flags, centers);
BUT, as it will run in a video stream, the number of clusters that i want will change as it picks up more points of interest. Is there a way to use radius from cluster centre to search instead of number of clusters?
thanks.