Abstract: Anti-occlusion is a very challenging research problem in video target tracking. During the target tracking process, when the target is partially occluded or completely occluded, the drift of the tracking model leads to the target loss. In order to solve this problem, this paper proposes a SiamVGG network target tracking algorithm with anti-occlusion mechanism. By analyzing the variation of the peak and connected domain of the output confidence map of the network, different tracking modes are set, which are normal tracking, partial occlusion, full occlusion and occlusion loss. Then we choose different tracking strategies according to different modes. Compared with other tracking algorithms, this algorithm adopts the SiamVGG network as the target tracking framework to analyze and correct the occlusion problem, effectively avoiding the target loss in the occlusion situation. Basing on the three benchmark datasets of OTB-50, OTB-100 and OTB-2013, we have verified the effectiveness and robustness of the algorithm proposed on the anti-occlusion problem.