In recent years, several visual tracking methods have applied multilayer convolutional features to correlation filters,but they mostly use fixed weights to fuse the multilayer response maps, which is difficult to adapt to various scenechanges. To address this problem, a robust tracking algorithm based on adaptive fusion of multilayer response maps isproposed. In this paper, we extract multilayer convolutional features from the target’s candidate area to improve thetracking robustness and the translation correlation filter is feed with CNN features extracted from each layer. Differentfrom previous methods, we proposed a fast covariance intersection algorithm to adaptive fuse the multilayer responsemaps. After the final target center position is determined, we adopted a 1D scale filter through multi-scale sampling withHOG features to handle large scale variations. Moreover, in order to solve the problem of tracking drifts due to thesevere occlusion and error accumulation, we present a new random update mechanism to update the translation filters.The experimental results on some challenging benchmark datasets show that the proposed algorithm achieves theoutstanding performance against the state-of-the-art tracking methods.
展开▼
