Motion recovery under perspective stereo vision

A problem of identifying motion of an object moving in the three-dimensional space is considered, assuming that the motion is described by a linear system with unknown constant motion parameters and that a single feature point on the object is perspectively observed by two cameras. We analyze the identifiability of motion parameters from the image data observed over an interval of time. It is shown that the parameters are identifiable generically. Moreover, the only case where the parameters can not be determined uniquely implies very restrictive motions, confined either in certain planes or lines in the three-dimensional space. Then by developing an experimental stereo vision system and using extended Kalman filter for motion recovery, we examined the validity experimentally.