Analytic methods for SAR image formation in the presence of noise and clutter

Abstract

Synthetic Aperture Radar (SAR) is a valuable imaging modality in civil and environmental monitoring, defense and homeland security applications. This thesis presents novel analytic and computationally efficient SAR image formation methods in the presence of noise and clutter using a priori models for target, clutter and noise.

The methods and algorithms derived in this thesis are extensively tested using high-fidelity simulated data and real SAR data.

The methods presented in this thesis have the advantages of computational efficiency, applicability to arbitrary imaging geometries and several different SAR modalities.

In the second part of the thesis, we investigate non-quadratic prior models to represent target scenes. Specifically, we consider edge-preserving prior models. First, we present a simultaneous analytic, image formation and edge detection method. Then, we formulate the SAR image reconstruction as non-quadratic optimization problems. We solve these optimization problems approximately with sequences of filtered-backprojection operators.

The first part of the thesis presents statistical SAR inversion methods to suppress noise and clutter using spatially-varying quadratic priors. We present a novel class of non-stationary stochastic processes, which we refer to as pseudo-stationary, to model radar targets and clutter. First, we develop analytic filtered-backprojection- and backprojection-filtering-type SAR inversion methods based on the minimum mean square error criterion when the target and the clutter are pseudo-stationary. Next, we develop an analytic inversion formula based on a best linear unbiased estimation criterion when the clutter is a pseudo-stationary process.