Skip to Main Content
We propose new features and a redundancy-free feature selection scheme for discriminating targets from clutter in high-resolution synthetic aperture radar imagery. These were found to be very useful for discriminating targets from clutter and well combined with various classical discriminative features. If a feature set with a larger dimension is used, its redundancy may increase. In order to reduce this redundancy, a rank-based feature selection scheme is devised. Our discriminating features and feature selection algorithm are evaluated using the moving and stationary target acquisition and recognition (MSTAR) data set.