Air-to-Air Detection and Tracking of Non-Cooperative UAVs for 3D Reconstruction

The monitoring and regulation of unauthorized, unresponsive, or uncommunicative non-cooperative UAVs is still one of the great challenges in airspace safety management. Therefore, in order to obtain the phenotypic information of these targets, we propose a 3D reconstruction system for non-cooperative UAVs based on air-to-air view image sequences. Specifically, the data-collecting UAV (pursuer) will detect, track, approach, and follow the non-cooperative UAV (target) at a fixed distance to automatically capture the image data. During the process, a lightweight object detection model and a Kalman filter-based tracking algorithm are used to recognize and continuously track the target UAV, and the boundary state constrained primitives and interval replanning are used to generate the accurate flight path, which in turn will control the pursuer UAV to approach the target UAV and follow it at a fixed distance. Finally, based on the implicit representation of the neural radiation field, the collected images will be used for 3D reconstruction and rendering of the non-cooperative UAV. In experiments, taking a quadrotor as the pursuer and a fixed-wing UAV as the target, we verify the feasibility of the proposed system in 3D reconstruction tasks.