Object Tracking Based on Saliency and Adaptive Background Constraint
Jing Wang, Weichao Huang · 2020
A target tracking algorithm with significant target indication and adaptive background constraint is proposed to solve the problem of tracking target lost caused by the disturbance of environment and the posturechange. Based on the particle filter tracking framework, the method weights the pixel features of the target region and the extended target region respectively, and constructs the target's indicative model. The background region is adaptively selected to constrain the tracking process according to the significance of the background region. The algorithm reduces the error in target matching and improves the tracking accuracy when the target is occluded or the attitude changes. Experimental results show that the algorithm has strong tracking robustness and high tracking accuracy.