A Vision-Based Ground Moving Target Tracking System for Quadrotor UAVs

Nowadays, unmanned aerial vehicles (UAVs) are playing their potential and role in various fields due to their unique advantages, and many tasks require UAVs to have the ability to stably track ground targets. In this paper, we design a vision-based tracking system, using which a quadrotor UAV equipped with a fixed monocular camera can track a moving target on the ground. The visual scheme includes a lightweight YOLOv3-tiny detector and a KCF tracker to provide ground target perception. And the relative position estimation between UAV and target is obtained according to the line-of-sight angle calculation and the Kalman Filter method. As for the real-time approach of the quad rotor UAV to the target, we adopt a PID velocity controller to provide continuous tracking capability. Finally, the results of simulation and flight experiments demonstrate that the designed system can enable the quadrotor UAV to perform stable tracking of a moving target on the ground.