CUEING, FEATURE DISCOVERY, AND ONE-CLASS LEARNING FOR SYNTHETIC-APERTURE RADAR AUTOMATIC TARGET RECOGNITION

The exquisite capabilities of biological neural systems for recognizin g target patterns subject to large variations have motivated us to inv estigate neurophysiologically-inspired techniques for automatic target recognition. This paper describes a modular multi-stage architecture for focus-of-attention cueing, feature discovery and extraction, and o ne-class pattern learning and identification in synthetic aperture rad ar imagery. To prescreen massive amounts of image data, we apply a foc us-of-attention algorithm using data skewness to extract man-made obje cts from natural clutter regions. We apply self-organizing feature dis covery algorithms that uniquely characterize targets in a reduced dime nsion space and use self-organizing one-class classifiers for learning target variations. We also develop a distance metric for partial obsc uration recognition. We present performance results using simulated SA R data and test for within-class generalization using nontrained targe ts including both in-the-clear and partially obscured examples. We tes t for between-class generalization using non-trained targets including both in-the-clear and partially obscured examples. We test for betwee n-class generalization using near-target data.