ONBOARD VISUAL TRACKING FOR UAV’S

Target tracking is one of the most common research themes in Computer vision. Ideally, a tracking algorithm will only once receive information about the target to be tracked and will be fast enough to identify the target in the remaining frames, including when its location changes substantially from one frame to another. In addition, if the target disappears from the area of interest, the algorithm should be able to re-identify the desired target. Target tracking was done using a drone with a Jetson TX2 computer onboard. The program runs at the drone level without the need for data processing on another device. Cameras were attached to the drones using a gimbal that maintains a fixed shooting angle. Target tracking was accomplished by placing it in the centre of the image with the drone constantly adjusting to keep the target properly framed. To start tracking, a human operator must fit the target he wishes to follow in a frame. The functionality of this system is excellent for remote monitoring of targets.