Adaptive context-aware correlation filter target tracking

Abstract

Abstract Aiming at the problem that the traditional correlation filter target tracking algorithm has low tracking accuracy under the conditions of fast motion, occlusion and complex background, an adaptive context-aware correlation filter target tracking algorithm is proposed in this paper. On the basis of the relevant filtering algorithm, the boundary effect and fixed learning rate brought by cyclic displacement are improved as the main purpose. Firstly, an adaptive sampling strategy based on the extreme value of the response graph is added to the context information in the training stage of the classifier. Then, A piecewise learning rate adjustment strategy is utilized to make the algorithm better adapt to the target change. Finally, the performance of the algorithm is verified by the standard data set. The experimental results show that the proposed algorithm improves the tracking accuracy of DCF and SAMF algorithm respectively. It not only has good robustness in the case of fast motion, occlusion, complex background, etc., but also can be integrated into most relevant filtering algorithms as a framework.