A Gimbal-Supported, Mono Camera, Relative Position Measurement System of a Visually Distinct Object for UAV Guidance

Abstract

This thesis describes a vision-based system for measuring distance and bearing between an unmanned aerial vehicle (UAV) and a ground object. The system is built of a single camera, mounted on a gimbal, which stabilizes camera attitude. The novelty of this system is that it uses a single camera and tracks a single object on the ground, hence requiring less equipment compared to other types of vision-based navigation systems. Vision system data estimates are fed into the controller to stabilize aircraft on a circular orbit and to maintain fully automated flight, without GPS data. The system can be used as a backup in the case of GPS signal loss. In that case, the UAV will orbit a target until the signal is restored or manual input is received. We describe the hardware chosen to solve this problem and the image-processing and rectification algorithms that determine the position of the target. The position data is used to control the autopilot which was optimized for circling the target. The real flight test results for the system, installed on a fixed-wing UAV, are presented.