Abstract:In order to improve the stability and tracking accuracy of the quadrotor UAV to the ground target, a tracking algorithm combining Tiny-YOLOV3 and Kalman filter is proposed. Firstly, the principle and network structure of Tiny-YOLOV3 are analyzed, and based on the target detection results of Tiny-YOLOV3, the mathematical model of target tracking system is established by combining the state of UAV and the geometric relationship of target; Then, the relative motion relationship of target is analyzed, and the kinetic model of target is established, considering that the detection result is greatly affected by interference, Kalman filter is applied to filter and predict the target trajectory, so as to improve the accuracy of target tracking. Finally, the UAV control law is designed according to the target trajectory information after Kalman filtering, at the same time, the yaw angle control of UAV is introduced to realize the UAV's stable tracking to the target. The simulation results show that the position tracking accuracy of UAV to target is within 0.5 m, the velocity tracking error is within 0.2 m/s, and the yaw angle tracking error is within 3 degrees, the tracking effect is good, which proves the effectiveness of the proposed algorithm.