Abstract:

A novel target tracking algorithm fused with the corner feature and the color feature is proposed to solve the poor stability and the antiblocking capability of the Mean-Shift algorithm. The invariance of Harris corner is used to solve the weak robustness of the Mean-Shift algorithm, and the kernel probability density estimation of the Mean-Shift algorithm is used to improve the ability of distinguishing target corners from background corners. Using a group of videos to test the proposed algorithm, the results show that the tracking stability and the anti-blocking capability of this algorithm are better than that of the Mean-Shift algorithm with single  corner feature or color feature.