A note on motion recovery of moving rigid body by perspective stereo vision

We consider a problem of recovering motion of rigid bodies moving in space under perspective observation. In particular, we consider the case of stereo vision, which provides us with much more information than the case of single camera. First, by an algebraic analysis, it is shown that the three-dimensional position of the body is generically identifiable. Then the position and the so-called motion parameters of the body are identified simultaneously by employing an extended Kalman filter. It is shown, by numerical simulations, that the recursive algorithm is effective even when the feature point is occluded.