Likelihood detection for nonfluctuating targets in ergodic K-clutter

Abstract

Non-Gaussian clutter distributions have been reported for high resolution radars operating over ocean surface. These observations have given rise to numerous non-Rayleigh clutter amplitude models; eg., log-normal, Weibull, and K. The authors extend these single point amplitude models to multipoint models, joint pdfs (jpdfs) of a vector observation with a prescribed amplitude pdf and covariance. This zero memory nonlinear transformation technique can be used to simulate ergodic WSS non-Gaussian random processes, as well to generate the jpdf of any N sample observation. Ergodicity is an important extension over the nonergodic SIRV model, in which the observation clutter amplitude pdf is not identifiable based on a single realization of any length. The authors utilize the jpdf to develop optimal likelihood ratio detectors for nonfluctuating target returns in K-clutter. The performance of the optimal detector is far superior to the matched filter.