Vision-Based Object Detection and Tracking of UAV with Unknown Camera Parameter

Vision-based object detection and tracking of Unmanned Aerial Vehicles (UAVs) has significant meaning for military or commercial applications in the fields of border patrol, target striking, criminal tracking, agricultural plant protection, and etc. To address the problem of UAV object detection and tracking of ground targets, this paper studies the vision-based object detection and target tracking control method for quad-rotor UAV with unknown camera parameter. Specifically, the YOLO v4 algorithm is utilized to achieve the onboard object detection of UAV and an optimized tracking controller based on LQR method is designed with camera parameters being unknown. Finally, a simulation environment is built based on the XTDrone simulation platform, and the effectiveness of the designed object detection and control algorithms is verified.