Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Lane Department of Computer Science and Electrical Engineering

Committee Chair

Matthew C. Valenti.


Swarmed Unmanned Aerial Vehicles (UAVs) with Automatic Target Recognition (ATR) technology are becoming an important element of electronic warfare. The advantages of UAVs over piloted vehicles have increased the need to develop robust and reliable ATR algorithms. Various issues like changing weather, camouflage, low contrast and resolution, clutter, inadequate databases place a limit on the performance capabilities of a typical ATR algorithm. In an effort to deal with these issues, a correlation-based algorithm is proposed in this thesis. This algorithm calculates the correlation between the input image and a target template which is created by projecting a 3-D model from the perspective of the UAV. The locations of correlation peaks are then declared to be the locations of the targets. We apply this algorithm to images with one object of a known class and move on to the more general case of images with an unknown number of targets from one or more classes. We provide an analysis of the performance of this correlation-based algorithm.;We compare the performance of the proposed correlation-based approach with that of a training-based approach. To provide a concrete example of an "off-the-shelf" training-based ATR algorithm, the open source IntelCV library was used. In the training-based method, a sample set is created and trained (Haar-like features are used for training) to produce results for comparison purpose. We further develop and analyze a method of correlating across multiple frames that have been preprocessed using the correlation-based approach. This method is shown to be useful in detecting true targets and suppressing false alarms in cases where a single image is not sufficient for classification.