Automatic Target Recognition (ATR) Research Validation of Synthetic Target Signatures (Topic I)

In the proposed program, TSC will evaluate candidate synthetic radar signature validation techniques. The objective is to provide the Air Force with a procedure for evaluating model-validation metrics for radar models used in developing Non-Cooperative Target Recognition (NCTR) techniques and algorithms. The methodology is to use the metric's classification performance against generic NCTR algorithms such as Adaptive Neural Nets and Bayesian classifiers to obtain a quantitative measure of each metric's performance as a prediction of NCTR effectiveness. The metric will be obtained with model fidelity as a parameter. Metrics will include statistical methods such as the Kolmogorov-Smirnov test and various correlation techniques. Preliminary validation analysis will be performed for representative targets at representative aspects using TSC's radar Imagery Generator (RIG) as the primary software platform. Extended analysis will be performed in the Phase II follow-on effort over a more extensive data base.