Target discrimination in complex synthetic aperture radar imagery

In previous work we demonstrated discrimination of natural background and manmade objects using low resolution synthetic aperture radar imagery is possible using singular value decomposition (SVD) methods applied to magnitude only imagery. For this work we use a subset of the multiscale autoregressive (MAR) model coefficients, and a subset of eigenvectors, all from complex imagery as features for target discrimination. The "optimal" set of features used to classify a region as "background clutter only" or "target region" is automatically chosen by a standard suboptimal feature selection algorithm. Additionally we use a support vector machine (SVM) for target discrimination and present preliminary results.