Clutter rank estimation for the multistage Wiener filter

Abstract

In airborne radar, reduced rank detection techniques are used when there are insufficient samples available for fully adaptive processing. This is the case in maritime radar, where the data can be both non-stationary and non-homogeneous. There are several approaches that have been proposed to address this problem. These include the single data set algorithms that eliminate the need for training data and reduced rank detectors such as principal components, cross-spectral metric and the multistage Wiener filter (MWF). This latter approach is superior to other rank reduction techniques in terms of computational efficiency and sample support requirements. In this paper, the authors propose an algorithm that determines the rank of the sea clutter and relates it to the number of stages in the MWF. The algorithm presented is formulated as a model order estimation problem that utilises the minimum description length (MDL). The authors present a computationally efficient implementation of the MDL and demonstrate its effectiveness using simulated data.