This forum is disabled, please visit https://forum.opencv.org
 | 1 | initial version |
I took part in a project where we detected cars from a flying camera. The assumption was that the camera was high enough to treat the ground as flat. In such a case, we were able to find homography between consecutive frames and stitch them together. Since two neighboring frames usually overlap substantially, we could detect differences in them AFTER stitching/overlaying. Actually we used more than just 2 frames, but if I remember well, less than 10 due to camera movement. Unfortunately, since the number of overlaying frames was not big, it was not possible to use some advanced background removal methods (like GMM for example) as they need several frames to learn the background. However, if your camera is not going to move too much, that could be possible.
