3D Motion Parameters Fusion Under a Multi-Vision Motion Capture System

Passive markers applied to motion capture system usually haven't any traits used to discriminate each other. Intricate human motion must lead to lose of markers in one binocular vision system. When the missing points reappear, identifying the marker belonged to which joints becomes a pivotal problem. Most available systems require manual steps to correct the tracking procedure. This work presents a novel approach based nearest neighbor method for identification such lost and reappearing marker. It combines an extended 3D Kalman filter and multi-trace data fusing technology, significant improving the accurately tracking rate. Experiments show that the proposed method can obtain the all markers' 3D motion parameters.