This proposed research is involved in designing systems capable of Automatic Target Recognition (ATR) in hostile environments. Special note is made of the unique aspects of atr in relation to the general problem of object recognition as studled in pattern recognition and computer vision. Research is then proposed that involves the development of an atr technique for use in ground-to-ground operation scenario using distributed sensor platforms, each containing a suite of complementary sensors. Research efforts to be emphasized in this proposed research project include: 1) develop a hierarchical 3-d multiview representation scheme to model targets so that they can be efpiclently recognized using sensor platforms dlstributed over a common geographical area; 2) develop and evaluate multisensor integration/pusion paradigms at the signal, pixel, peature and symbol level; 3) develop a stand-alone hierarchical multi-processor computer system to support real-time operation of an intelligent mobile sensor platporm. In thePhase I effort, simulations are used to demonstrate how this technique can improve overall target recognition accuracy as compared to the use of just a single sensor or a suite of sensors mounted on a single platform. The simulations include a variety of ground-based targets that are presented in a number of configurations and under dlfferent operating conditions. Anticipated benefits /potential commercial applications - the feasibility study of the proposed Phase I effort is to provide a solution to the key issues for an effective automatic target recognition system for military applications. The research results could be adapted for use in a variety of commercial applications.
