Parallel Implementation of Covariance Tracking with Efficient Model Update

Most of the appearance based tracking algorithm developed in the recent past are characterized by high computation-intensive operations and demands high memory and performance requirements. These appearance based model are also highly sensible to the variation of extrinsic and intern-sic parameter of the feature. In order to track the object under the variation of intern-sic and extern-sic factor , a new model update approach is developed and implemented using a thread level parallelism. Furthermore Particle filter is added to this current method to better handle the back ground clutter, as well as the temporary occlusion. Parallelized implementation achieves significant speedup, and meets the target frame rate under various configurations. Simulation shows that the current parallel method is robust and very effective for the object tracking.