python - How to find multiple instances of my training image with OpenCV and SIFT -
so far i've been able detect key points training image within query image accurately using bfmatcher. however, have dozens of instance of training image (a bee) occurring within query image (the frame of beehive). possible use sift multiple independent instances of image?
ideally, i'm hoping point wit above image can match 4 key points query image against multiple independent bees.
i don't know opencv enough know if offers routines help. said, math diy isn't bad if have numerical library svd e.g. numpy. in lowe's 2004 paper, match determined ratio of 2 closest matches. if compute keypoints in 'query image' , apply criteria return multiple results if present (any 2 keypoints in database same ratio +/- epsilon match). however, can result in spurious matches. so, in addition basic test lowe paper, ransac can used discard candidate matches not consistent homography between 'training image' , 'query image'. see 'distinctive image features scale-invariant keypoints' david lowe , google "ransac homography" details. i'm not sure if desire match 4 keypoints critical. technique i'm referring use 4 points calculate candidate homography @ each ransac iteration, total number match can >4 if happens match , consistent homography. using 4 points only possible using ratio test, not possible using ransac.
Comments
Post a Comment