Designing Empirical Lab Experiments for SAR-ATR

Automatic target recognition (ATR) experiments often rely on simulated data from high-fidelity digital models or from highly-accurate synthetic aperture radar (ISAR) measurements of real models in an anechoic chamber. Both approaches have high costs that limit opportunity to perform rapid or ad hoc experiments. Here, we present recent work towards ad hoc ATR experimentation with 1:144-scale model commercial aircraft. Although relatively inexpensive and fast, measurements at Ka-band with a non-anechoic test site and quasi near-field ISAR configuration pose challenges when designing an ATR experiment. The local measurement technique, data processing, ATR feature extraction, and classification method are described with emphasis on the various tradeoffs at each stage.