Multi-target Collaborative Identification and Tracking of UAV Swarm with Passive Sensors

The environment perception in a complex battlefield by the swarm of unmanned aerial vehicles (UAV) have increased considerably in the last few years, which focuses on the detection, identification and tracking of maneuver targets with arbitrary number. This paper proposed a passive collaborative multi-target identification and tracking method for the surveillance of UAV swarm to ground maneuver targets, which based the YOLOX architecture and Labeled Multi-Bernoulli (LMB) filtering algorithm. The vision process of each sensor or camera segmented and identified the targets with the neural network of YOLOX and calculated the two measurement angles by the affine transformation. Considering the perspective divergence observed by the different passive sensors of UAVs, the trajectories association of multi-target from the sensors is considered with Gaussian Mixture LMB (GM-LMB) filter which could decrease the calculation complexity and relevancy uncertainty. The results of mathematical simulation indicate that the GM-LMB filter could compensate the tracking position error compared with Gaussian Mixture Probability Hypothesis Density (GM-PHD) filter.