The Object Tracking Based on Integral Covariance Matrix

The object tracking by using single feature is possible to generate errors and easy to lose the target if the illumination and object size scale are changed. We propose a particle-filter-object-tracking algorithm. The proposed algorithm is based on a covariance region descriptor (CRD). The CRD can fuse different features of a targeted object region while handling various complex backgrounds. Hence, the robustness of tracking algorithm is achieved. Moreover, the integral covariance matrix computation is an extension to Bayesian tracking framework, which makes the tracking more efficiency and for handling high performance tracking in real-time. The comparative experiments show that the proposed algorithm is more robust and its efficiency of computation of tracking is higher performed than the one uses traditional the object tracking algorithm with only consideration of single feature.