In this thesis we present a complete terrain following framework composed of an experimental UAV (Unmanned Aerial Vehicle) platform specifically designed for our research, a height estimation algorithm, a reliability classifier and a proportional height controller. We realized the astounding growth of UAV usage, notably in mapping applications, would soon require the creation of new technologies to enable these systems to perceive their surroundings. Specifically we chose to tackle the terrain following problem, as it is probably a relevant one for autonomous missions, yet still unresolved for consumer available systems. Virtually every mapping aircraft carries a camera, therefore we chose to exploit this in order to use presently available hardware to extract the height information toward performing terrain following flights. Our methodology consists of using optical flow to track features from videos obtained by UAV as well as its motion information to estimate the flying height. However, since our vehicle is a multirotor, it is susceptible to winds and other disturbances that might induce yaw, which we did not address in this study. Therefore, to determine if our height estimation is reliable, we trained a decision tree that takes the optical flow information as input and classifies whether the output is trustworthy or not. In addition, we also propose a waypoint based height controller to guide the aircraft to the desired height. Our classifier achieved accuracies of 80% for positives and 90% for negatives, while our height estimation algorithm was more accurate than other alternatives, but less precise, which is consequence of lower noise tolerance. Our controller was tested in simulations and even when subject to 10% noise it worked successfully..

Prof. Luiz Chaimowicz – Advisor (DCC – UFMG)
Prof. Erickson Rangel do Nascimento – Co-advisor (DCC – UFMG)
Prof. Douglas Guimarães Macharet (DCC – UFMG)
Dr. Gustavo Medeiros Freitas (ITV)
Prof. Vilar Fiuza da Camara Neto (DE – FUCAPI)