I know these cameras are not designed for strict motion capture other than head tracking, but I'm using them on a project where the users hands have to be tracked and their motions replicated in real time, in a 3D scene.
I assign markers to "bones" based on their spatial coordinates on the tracked data (i.e. on a top camera the left-most marker being tracked corresponds to the left hand, on a side camera, the left most marker belongs to the distal phalanx, etc.). The problem is, the markers being passive, these assignments based on the coordinates are very weak in terms of reliability, at any given time you might receive erroneous info due to some markers being lost by the cameras, or some going cross some coordinates that "don't belong to them".
My question is more to seek for any advice on filtering this kind of info and algorithms to track the motion based on previous positions (i.e. a system that actually "follows" the marker over time through its smooth movements) than actual technical support...
Any advice on this?
PD. Sorry if I didn't explain the problem properly, English is not my first language.
1 post • Page 1 of 1