Lucas-Kanade check next keypoints

asked 2018-02-13 11:38:44 -0500

updated 2018-02-13 11:53:06 -0500


under the example of the lucas-kanade algorithm is written "(This code doesn't check how correct are the next keypoints. So even if any feature point disappears in image, there is a chance that optical flow finds the next point which may look close to it. So actually for a robust tracking, corner points should be detected in particular intervals.)"

Is there another option to check, if the next keypoints in the next image make sense? For example: there is a video on this page, where you can see a pencil on the table with a tracked point. After the guy moves the box, this point is then tracked on the long side of the box. But in my opinion, this tracked point of the pencil is lost, so it´s not plausible, that this point is now attached to the box...Does anyone know how to check this?

greets Matthias

edit retag flag offensive close merge delete