Abstract:  Aiming at the problem of poor target tracking stability caused by single-feature target-related filtering algorithms due to illumination changes, target occlusion, low resolution, and motion blur, this paper proposes a tracking algorithm that adaptively fuses multiple features . Based on the FDSST algorithm, the algorithm in this paper adaptively fuses gradient histogram features, color name features and grayscale features to enhance the expressive ability of features; proposes an occlusion judgment strategy, which can effectively determine the occlusion phenomenon of targets in the tracking process; introduce targets The relocation mechanism can relocate the target position when the target is blocked or interfered, effectively suppressing the occurrence of tracking drift. Finally, this paper selects OTB50 and OTB100 as the experimental data sets, and compares the performance of the algorithm proposed in this paper with the selected six mainstream algorithms. On the data sets OTB50 and OTB100, the accuracy of the algorithm in this paper is 89.6% and 90.9%, which are 7.9% and 1.8% higher than the second-ranked algorithm respectively. The experimental results show that the algorithm in this paper has high stability and accuracy under the conditions of illumination changes, motion blur and target occlusion; it is better than the other six algorithms in success rate and tracking accuracy.