LAUSR

Unmanned aerial vehicle (UAV) swarms show a broad application prospect in the future. For most individuals of UAV swarms, their perception systems consist of onboard camera, gimbal and odometry. It is necessary to efficiently obtain accurate extrinsic parameters of camera-gimbal-odometry systems for UAV swarms. Traditional extrinsic calibration methods for camera-gimbal-odometry system often need manual assistances. For a large number of individuals of UAV swarms, it is time- and labor-consuming to complete the calibration using traditional methods. To tackle this problem, we propose an online extrinsic calibration algorithm and design a parallel procedure to realize an efficient calibration for UAV swarms. Without any dependence on manual operation and communication with ground station, the calibration for each individual is completed on onboard processor. Instead of manual labelling, a fast and accurate detection module is firstly built for samples auto-labeling. Facing the challenge of limited computing power of onboard processor, our auto-labeling algorithm shows a strong real-time capability and robustness. Then, an optimization module is developed to iteratively refine the extrinsic. Using multiple fixed-wing UAVs, online calibration experiments were conducted in larger-scale outdoor, and the results validated the feasibility of our method. Compared with some state-of-the-arts, the proposed extrinsic calibration method showed superior performance in terms of efficiency, accuracy and robustness.