If you wish to better control the number of detected points, a special subclass of the cv::FeatureDetector class, called cv::DynamicAdaptedFeatureDetector, is available. This allows you to specify the number of interest points that can be detected as an interval. In the case of the FAST feature detector, this is used as follows:
cv::DynamicAdaptedFeatureDetector fastD(
new cv::FastAdjuster(40), // the feature detector
150, // min number of features
200, // max number of features
50); // max number of iterations
fastD.detect(image,keypoints); // detect points
The interest points will then be iteratively detected. After each iteration, the number of detected points are checked and the detector threshold is adjusted ...
