Anti-Occlusion Object Tracking Algorithm Based on Kalman Filter

When Unmanned Aerial Vehicle (UAV) performs object tracking tasks, it faces problems such as target out-of-view and occlusion, which have a great impact on the tracking performance. In this paper, an anti-occlusion object tracker based on a Kalman filter is proposed, which can effectively deal with the occlusion problem of objects. Firstly, a high confidence measure was proposed to detect tracking anomalies, and Kalman filter was introduced to predict the object motion trajectory and generate object candidate regions. Secondly, the historical target template is used to correlate the target candidate area to obtain the relevance response map, and the target relocation is realized according to the relevance response map. The proposed algorithm is compared with other tracking algorithms on two UAV benchmark datasets, and the results show that the proposed algorithm has better tracking performance.