A Collaborative Relay Tracking Method Based on Information Fusion for UAVs

In this article, a collaborative relay tracking method based on information fusion is proposed for unmanned aerial vehicles (UAVs). Target matching is performed by combining the target's spatiotemporal information with its appearance and other auxiliary cues. Only the motion information and a compressed image of the target are used for target handover. Specifically, to determine the spatiotemporal transfer probability of the target after entering the blind area of visual field, this article proposes a spatiotemporal correlation model construction method to predict the spatiotemporal similarity. Besides, Siam-LGNet is proposed in this article for target appearance matching to overcome the effects of various view angles, rotations, scale changes, and other factors on the images. This effectively ensures the accuracy of target matching and lowers the model complexity. In addition, similarities obtained from multiple cues are fused by D–S evidence theory to compensate for the uncertainty of single cue. Finally, the effectiveness of the proposed method is verified in simulation experiments and physical experiments. This article fills a gap in the field of UAVs relay tracking. The proposed scheme is scalable and lays the foundation for subsequent research on UAVs relay tracking problems.