Maritime Target Detection in Non-homogeneous Sea Clutter Environments based on Single- and Multi-polarization Synthetic Aperture Radar Data

Abstract

This thesis discusses the subject of maritime target detection based on single- and multi-polarization synthetic aperture radar (SAR) data. The primary objective is to develop an automatic and effective target detection algorithm, which is able to provide robust performance for an operational maritime surveillance system under various circumstances.

There are two frequently encountered major detection issues in practice. Firstly, in multiple-target situations, the local reference sea clutter is often contaminated by interfering targets. The outcome is known as the capture effect. Secondly, in non-homogeneous environments, sea surface transitions between regions with different radar backscattering properties are usually observed in conjunction with various meteorological and oceanographic phenomena. The result is recognized as the clutter edge effect. Both effects inevitably lead to inaccurate parameter estimation and deceptive statistical modeling, thus causing severe degradation of the detection performance.

In this study, the conventional constant false alarm rate (CFAR) detector setup is adopted in the algorithm. The rigorous statistical analysis using truncated statistics (TS) provides improved background clutter modeling results, while a corresponding TS-based CFAR (TS-CFAR) detector is first proposed to suppress the capture effect and ameliorate detection performance. Moreover, an automatic image segmentation stage with suitable statistics is then adopted in the detection scheme to address the clutter edge issue simultaneously. It provides a comprehensive statistical analysis of the non-homogeneous background clutter and obtains the local contextual information for the subsequent CFAR detection stage.

The end product is a segmentation based CFAR detection algorithm using TS. Single-look intensity (SLI) and multi-look intensity (MLI) SAR imagery are mainly targeted in this study for directly supporting operational applications. It has been demonstrated on several real SAR images, that the proposed algorithm is able to adapt to various contaminated non-homogeneous environments, provide improved local background clutter modeling, and deliver robust detection performance.