Neuromorphic Optical Signal Processing and Image Understanding for Automated Target Recognition

Abstract : The goal of research described in this report is study of computation and learning in neural net models and demonstration of their utility in image understanding and neuromorphic information processing systems for remote sensing and target identification. The approach to achieving this goal has two facets. One is combining innovative architectures and methodologies with suitable algorithms to exploit existing and emerging photonic technology in the implementation of large-scale neurocomputers for use in: (a) the study of complex self-organizing and learning systems, (b) fast solution of optimization problems, (c) feature extraction, (formation of object representation), and (d) pattern recognition. The second facet of the approach is to demonstrate and assess the capabilities of neuromorphic processing in solution of selected inverse-scattering and recognition problems. The problem we have chosen to study as test bed for our work is that of automated radar target recognition because of our existing capabilities and expertise in this area.