Hello, I am currently using the sparse optical flow algorithm and I have a few questions.

As a an object enters the FOV of the camera, its velocity is tracked as near zero. This is because as the object moves into view, its "center" is staying apparently near the edge until the full object is in view. Is there a way to mitigate this without throwing away the initial velocity data?

Secondly, is there a way to pull statistics/certainty for the calculations the algorithm runs?

And, is there a way to associate object IDs with certain objects in view? I'm getting hundreds or thousands of distinct IDs, and have no way of showing which object is providing that movement.

Thank you.