2D-2D pose estimation?
I have used HoughCircles to find the centre points (and radii) of a group of circles in an image. I now have another image, which might or might not contain the same group of circles, but possibly rotated and/or scaled, and with the possibility of some noise (i.e. extra circles, and/or missing circles, and some circles could be slightly moved).
So I have two sets of x,y locations and radii. My question is, can OpenCV help me 'match' the points in one image with the points in another? I can't use any of the standard 'findHomography' functions or similar, as I don't know which pixel pairs match. I have tried using SIFT/SURF/ORB on images containing just the centre locations, but they almost always try to match incorrect pairs.
To give an idea of numbers: There could be between 6 and 30 circles in each image, i.e. I'm trying to match between 6-30 pairs of 'points' (circles centres) between the two images.
Any ideas would be gratefully received - thanks in advance.
Please always post example images with your question! It's then much easier to understand the problem. Depending on your problem (e.g. number of circles) the RANSAC-Algorithm could help you.
Thanks for the info - I will look into RANSAC now. Sorry I didn't provide sample data - in this case it's really just 20 or so white points in a black space, and I'll post images if I need follow-up. Thanks again,