EFFICIENT ATR USING COMPRESSION

We examine various model-based automatic target recognition (MBATR) cl assifiers to investigate the utility of model-catalog compression real ized via signal-vector quantization (VQ) and feature extraction. We sp ecifically investigate the impact of various compression rates and com mon automatic target recognition (ATR) scenario variations such as noi se and occlusion through simulations on high-range resolution (HRR) ra dar and synthetic aperture radar (SAR) data. For this data, we show th at significant computational savings are possible for modest decreases in classification performance.