The Clutter SIRP and Gaussian Models: a Brief Overview and a Comparison

At low grazing angle and/or high resolution radar, sea clutter is not Gaussian due to the apparition of spikes. It is then generally seen as a spherically invariant random process (SIRP). When the covariance matrix is assumed to be known, the Neyman-Pearson optimal receiver can be derived. Nevertheless, in order to implement it, the SIRP characteristic probablity density function (PDF), that is to say the texture PDF, has to be known. Its asymptotical form, which does not depend on the texture PDF, is then often prefered. We show in this paper that the asymptotic detection in a SIRP is equivalent to the generalized likelihood ratio test (GLRT) in a Gaussian correlated clutter of unknown level, and discuss about the interest of SIRP model and optimality.