PROJECT TITLE :

Performance prediction of quantized SAR ATR algorithms

ABSTRACT:

Automatic target recognition (ATR) of artificial aperture radar (SAR) target chips may be a troublesome drawback, sophisticated by the quantity of nuisance parameters typically present in SAR imagery. This conjointly complicates performance analysis as a result of fully sampling the house of nuisance parameters in an evaluation dataset is intractable. ATR algorithms that 1st quantize pixel intensity values are shown to be effective for SAR ATR because of hypothetically reducing the sensitivity to these nuisance parameters. Here we tend to study the performance of two such algorithms, multinomial pattern matching and quantized grayscale matching, and compare them with the traditional mean squared error (MSE) template matching based mostly classification approach. Our approach is to approximate the decision statistic of every algorithm as a Gaussian random variable (RV) parameterized by the noise power, or alternatively signal-to-noise ratio (SNR). This allows the analytic prediction of algorithm performance when the noise process of take a look at pictures differs from that of the dataset used to train every algorithm, while not having to depend on expensive empirical simulation. We verify our leads to simulations utilizing the AFRL "Civilian Vehicle" dataset.