Classification Schemes for the Radar Reference Window: Design and Comparisons

TL;DR

This paper proposes classification schemes for radar reference window data to identify homogeneous clutter regions, improving clutter mapping and aiding more reliable detection by analyzing statistical properties and scenario-specific hypothesis tests.

Contribution

It introduces novel classification architectures based on multiple hypothesis testing and model order selection for radar clutter analysis, enhancing scenario recognition and clutter mapping.

Findings

Classification architectures effectively identify homogeneous clutter regions.

Proposed methods improve scenario recognition accuracy.

Clutter maps assist in more reliable radar detection.

Abstract

In this paper, we address the problem of classifying data within the radar reference window in terms of statistical properties. Specifically, we partition these data into statistically homogeneous subsets by identifying possible clutter power variations with respect to the cells under test (accounting for possible range-spread targets) and/or clutter edges. To this end, we consider different situations of practical interest and formulate the classification problem as multiple hypothesis tests comprising several models for the operating scenario. Then, we solve the hypothesis testing problems by resorting to suitable approximations of the model order selection rules due to the intractable mathematics associated with the maximum likelihood estimation of some parameters. Remarkably, the classification results provided by the proposed architectures represent an advanced clutter map since, besides the estimation of the clutter parameters, they contain a clustering of the range bins in terms of homogeneous subsets. In fact, such information can drive the conventional detectors towards more reliable estimates of the clutter covariance matrix according to the position of the cells under test. The performance analysis confirms that the conceived architectures represent a viable means to recognize the scenario wherein the radar is operating at least for the considered simulation parameters.