Vision Based Control of Autonomous UAV

Abstract: This master thesis investigates the problem of making an unmanned aerial vehicle(uav) follow a person or a group of persons while keeping a fixed distance tothe chosen target. The purpose of this thesis is to give a proof of concept prototypeof how such a system would work to achieve that task. The main problemconsists of controlling the uav based on visual input from a camera. With thehelp of a visual object detection and tracking system, the image coordinates ofthe targets can be found. An algorithm was developed to calculate the target’sworld position based on its image coordinates and the world position and orientationof the uav. A control system was implemented that uses that uses thepositional information to set the velocity of the uav, if its position needs to bechanged. Several strategies for handling groups of targets were investigated. Inaddition a simulator was developed that can be used to simulate the image coordinatesof a target when the world position of the target is known. The systemwas tested during live flights, using a high precision motion capture system forreference. The results were mainly positive in showing proof of concept and evenshowing a relatively high level of precision.