Hi,
I’ve been teaching motion tracking for a while and one question that pop a lot is how students can identify which keyframes should they use.
I tell them that markers should be distributed in the shot, tracking foreground and background objects and they should look for perspective shift on the shots.
Because the answer is too close to “it depends”, I tried to get a more correct answer and did some tests looking into the graph mode on the motion tracking editor. I draw this conclusion that can be totally wrong, but I couldn’t find any confirmation anywhere so a “peer review” is welcome and “check this research paper about the algorithm” would be totally awesome.
- each tracker chooses an interval of possible locations with a min and max value (red and green lines)
- when the camera is solving the algortihm uses the averages of at least 8 min values and 8 max values and averages them.
- students should look for the biggest amplitudes between red and green lines, because they cover more ground, in order to identify the best candidates for solving.
Can anyone confirm if my conclusion is right (even if the steps are not)?
Thanks.