PROJECT ABSTRACT



Viola Jones face detection, Camshift, Haar-Cascades, quadrilateral transformation, fiducial marker recognition and Fisherfaces are algorithms and techniques commonly used for object detection, recognition and tracking. This research evaluates the methodologies of these algorithms in order to determine their relative strengths, weaknesses and suitability for object detection, recognition and tracking using a quadrotor. The capabilities of each algorithm are compared to one another in terms of their effective operational distances, effective detection angles, accuracy and tracking performance. A system is implemented that allows the drone to autonomously track and follow various targets whilst in flight. This system evaluated the feasibility of using a quadrotor to provide autonomous surveillance over large areas of land where traditional surveillance techniques cannot be reasonably implemented. The system used the Parrot AR Drone to perform testing owing to the drone's on board available sensors that facilitated the implementation of computer vision algorithms. The system was successfully able evaluate each algorithm and demonstrate that a micro UAV like the Parrot AR Drone is capable of performing autonomous detection, recognition and tracking during flight.