Dynamic target tracking with multi-feature covariance based on Kalman filter predictor

Aimed at target tracking in the video image sequences, this paper introduces a dynamic objects tracking algorithm based on the combination of Kalman prediction and covariance module updating. Via kalman prediction, the getting of the dynamic interesting regions in the next frame of the image sequences, an operation which facilitates the realization of the real-time target localization, can be realized. Meanwhile, the updating of the target covariance matrix and the prediction of the target marching regions also improve the disturbance rejection performance, robustness of the whole tracking algorithm. Experiments results show that the algorithm introduced in this paper is much better than the covariance tracking algorithm based on static template in the tracking performance and real-time character.