OpenCV BFMatcher - Ignore False Positives - opencv

Question

Can someone describe a good process for ignoring false positives from BFMatcher?
I define an image to find in the scene, use SiftFeatureDetector, SiftDescriptorExtractor, then use a BFMatcher. When searching for the correct marker, I find the matches no problem, but I want to make my code more robust to false positives.
//Detect keypoints using ORB Detector
SiftFeatureDetector detector;
vector<KeyPoint> keypoints1, keypoints2;
detector.detect(im1, keypoints1);
detector.detect(im2, keypoints2);
//Draw keypoints on images
Mat display1, display2;
drawKeypoints(im1, keypoints1, display1, Scalar(0,0,255));
drawKeypoints(im2, keypoints2, display2, Scalar(0,0,255));
//Extract descriptors
SiftDescriptorExtractor extractor;
Mat descriptors1, descriptors2;
extractor.compute( im1, keypoints1, descriptors1 );
extractor.compute( im2, keypoints2, descriptors2 );
BFMatcher matcher(NORM_L1, true);
vector<DMatch> matches;
matcher.match(descriptors1, descriptors2, matches);
I tried to filter out false positives by skip:
if (matches.size() < 50) {
//false positive - skip
} else {
//perform actions
}
But this is not at all robust. I think I saw a few articles on people using radius matchers, but I couldn't find a good description of using a radius match with Brute Force. I checked the documentation: http://docs.opencv.org/modules/features2d/doc/common_interfaces_of_descriptor_matchers.html , but it was extremely clear to me how I decide what is a good min_dist/max_dist for this application?
I'm sure this is a pretty simple answer for some of you - your help is greatly appreciated!

Answer 1

You need to filter on the distance of the matche.
Take care that distance is dependent of the norm you choose in BFMatcher.
Here an exemple from openCV samples :
double min_dist = 100;
for( int i = 0; i < descriptors_object.rows; i++ )
{ double dist = matches[i].distance;
if( dist < min_dist ) min_dist = dist;
}
/** Keep only "good" matches (i.e. whose distance is less than 3*min_dist ) **/
std::vector< DMatch > good_matches;
for( int i = 0; i < descriptors_object.rows; i++ )
{ if( matches[i].distance < 3*min_dist )
{ good_matches.push_back( matches[i]); }
}