Implementation of an Onboard Visual Tracking System with Small Unmanned Aerial Vehicle (UAV)
This addresses the need for autonomous object tracking in UAV applications, but it is incremental as it builds on existing visual tracking methods with specific adaptations for UAV constraints.
The paper tackles the problem of enabling real-time visual tracking on small UAVs by developing a computationally efficient system that is invariant to lighting changes and rotation, with performance verified in lab and flight tests showing it can track diverse objects in real-time.
This paper presents a visual tracking system that is capable or running real time on-board a small UAV (Unmanned Aerial Vehicle). The tracking system is computationally efficient and invariant to lighting changes and rotation of the object or the camera. Detection and tracking is autonomously carried out on the payload computer and there are two different methods for creation of the image patches. The first method starts detecting and tracking using a stored image patch created prior to flight with previous flight data. The second method allows the operator on the ground to select the interest object for the UAV to track. The tracking system is capable of re-detecting the object of interest in the events of tracking failure. Performance of the tracking system was verified both in the lab and during actual flights of the UAV. Results show that the system can run on-board and track a diverse set of objects in real time.