Radar Signal/Image Processing Enhancements using Alpha-Stable Techniques

Abstract

Abstract : In conventional radar signal and image processing, the background clutter and noise are assumed to follow the Gaussian model. Under this assumption, it has been shown that the conventional matched filter is optimal in target detection. However, recent research has found that many nonhomogeneous types of clutter and noise, such as clutter, do not fit the Gaussian model well because of impulsive outliers or the so called sea spike. These types of clutter and noise lend themselves to a heavy tail in amplitude distribution; consequently, the conventional matched filter does not perform well. Most recent research has shown that the alpha-stable model is a better model. The alpha-stable model is a natural extension of the Gaussian model, and most radar clutter is modeled well by the alpha-stable statistics. A robust family of alpha-stable matched filters is a natural extension of the conventional matched filter. An optimal alpha-stable matched filter extracted from that family is being developed in a simple closed form. This optimal alpha-stable matched filter significantly improves target detection in both real clutter data and simulated data. Moreover, the alpha-stable matched filter is computationally efficient. It can be applied in wide varieties of radar signal and image processing.