Motion tracking strategies for flying robots

Abstract

The principal goal of this thesis is to see whether it is possible to achieve real-time motion detection and tracking with a flying robot using computer vision. In order to achieve this goal, two separated but dependant works have been be done. On one hand, a good theoretical analysis on the different techniques of a tracking process has been studied. On the other hand, a software prototype containing the chosen tactics has been implemented. The strategy for the development was to follow an exploratory path by using ViBe, a background subtraction algorithm, as a starting point, adapting and expending it to fit the requirements. The conclusion of this work is that the modifications made on ViBe significantly improve the results for a moving camera. The tests, made on video sequences of 320 × 200 pixels, achieve near real-time on the development platform.