Research on Mean Shift Tracking Algorithm Based on Significant Features and Template Updates
Hui Wang, Zhang Xue, Lijun Yu, Xueying Wang · 2018
Based on the problem of inaccurate positioning of the target in tracking dynamically by the traditional Mean Shift algorithm due to background disturbance and lack of template update mechanism, two improvement measures are presented in this paper. On the one hand, the prior information of the target is used to generate the visual saliency map, and the central reinforcement-peripheral weakening mechanism is applied to further weaken the influence of the background area, and the significant feature of the target is extracted to replace the color feature; On the other hand, in the tracking process, a decision method based on the variance size is used to dynamically determine the target model update strategy. The experimental results show that the proposed algorithm can accurately track the moving target in the video streams.