Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Lane Department of Computer Science and Electrical Engineering

Committee Chair

Natalia A. Schmid.


Automatic Target Recognition (ATR) is computer processing of images or signals acquired by sensors with the purpose to identify objects of interest (targets). This technology is a critical element for surveillance missions. Over the past several years there has been an increasing trend towards fielding swarms of unattended aerial vehicles (UAVs) operating as sensor networks in the air. This trend offers opportunities of integration ATR systems with a UAV-based sensor network to improve the recognition performance. This dissertation addresses some of design issues of ATR systems, explores recognition capabilities of sensor networks in the presence of various distortions and analyzes the limiting recognition performance of sensor networks.;We assume that each UAV is equipped with an optical camera. A model based recognition method for single and multiple frames is introduced. A complete ATR system, including detection, segmentation, recognition and clutter rejection, is designed and tested using synthetic and realistic images. The effects of environmental conditions on target recognition are also investigated.;To analyze and predict ATR performance of a recognition sensor network, a general methodology from information theory view point is used. Given the encoding method, the recognition system is analyzed using a recognition channel. The concepts of recognition capacity, error exponents and probability of outage are defined and derived for a PCA-based ATR system. Both the case of a single encoded image and the case of encoded correlated multiple frames are analyzed. Numerical evaluations are performed. Finally we discuss the joint recognition and communication problems. Three scenarios of a two node recognition sensor network are analyzed. The communication and recognition performances for each scenario are evaluated numerically.