Date of Graduation
Statler College of Engineering and Mineral Resources
Mechanical and Aerospace Engineering
Marcello R Napolitano
This research presents experimental results for the application of Machine Vision (MV) techniques to address the problem of target detection and tracking. The main objective is the design of a prototype "UAV" surveillance environment to emulate real-life conditions. The model environment for this experiment consists of a target simulated by a small electric train system, located at "ground" level, and a MV camera mounted on a motion-based apparatus located directly above the model setup. This system is meant to be a non-flying mockup of an aerial robot retrofitted with a MV sensor. Therefore, the final design is a two degree-of-freedom gantry simulating aircraft motions above the "ground" level at a constant altitude. On the "ground" level, the design of the landscape is an attempt to achieve a realistic natural landscape within a laboratory setting. Therefore, the scenery consists of small scale trees, bushes, a mountain, and a tunnel system within a 914 mm by 1066 mm boundary. To detect and track the moving train, MV algorithms are implemented in a Matlab/SimulinkRTM based simulation environment. Specifically, image pre-processing techniques and circle detection algorithms are implemented to detect and identify the chimney stack on the train engine. The circle detection algorithms analyzed in this research effort consists of a least squares based method and the Hough transform (HT) method for circle detection. The experimental results will show that the solution to the target detection problem could produce a positive detection rate of 90% during each simulation while utilizing only 56% of the input image. Tracking and timing data also shows that the least squares based target detection method performs substantially better then the HT method. This is evident from the result of using a 1--2 Hz frequency update rate for the SimulinkRTM scheme which is acceptable, in some cases, for use in navigation for a UAV performing scouting and reconnaissance missions. The development of vision-based control strategies, similar to the approach presented in this research, allows UAVs to participate in complex missions involving autonomous target tracking.
Effland, Joshua Patrick, "Application of Machine Vision in UAVs for Autonomous Target Tracking" (2008). Graduate Theses, Dissertations, and Problem Reports. 4367.