Development of Vision-Based Human Tracking for Drone’s Gimbal

The use of Unmanned Aerial Vehicles (UAVs) with different features and for a variety of applications has grown significantly. Tracking generic targets, especially human, using the UAV's camera is one of the most active and demanding fields in this area. In this paper we implement two vision-based tracking algorithms to track a human by using a 2D gimbal which can be mounted on UAVs. To ensure smooth movements and reduce the effect of common jumps on the trackers output, the gimbal motion control system is equipped with a Kalman filter followed by a proportional-derivative (PD) controller. Various experimental tests have been designed and implemented to track a human. The evaluation results show success in tracking the high speed movements with one of the algorithms and high accuracy in tracking the challenging movements in the other algorithm. Also in both methods, the tracking computation time is short enough and suitable for real-time implementation. The favorable performance of both algorithms indicate the ability of designed system to be implemented on the UAVs for practical applications.