Multiscan Recursive Bayesian Method for Parameter Estimation of Spatially-Varying Sea Clutter Models

Abstract

Large-scene sea clutter exhibits spatially-varying power and non-Gaussianity due to varying grazing angle and sea state, which leads to the lack of data with consistent statistical property, called spatially small-sample (SSS) problem. Therefore, compound-Gaussian model with spatially-varying texture is introduced to characterize large-scene sea clutter. However, existing methods of sea clutter parameter estimation, including moment-based, maximum likelihood and explicit bipercentile estimators, lose their estimation efficiencies with the SSS problem. In this study, a multiscan recursive Bayesian (MRSB) method is proposed to deal with the estimation problem of spatially-varying sea clutter parameters. This method is constructed by transforming scanning data in frames to the prior information of parameters in order to cover the shortage of sample number. The MRSB method not only improves the parameter estimation property of spatially-varying sea clutter due to the performance assessment with clutter data, but also starts a new mode of radar, that is, radar remembers all information and no data.