Automatic Parameter Tuning in Parametric and Regularization TomoSAR Focusing Methods

Abstract

In the context of direction-of-arrival, super resolution focusing techniques like the parametric method multiple signal classification (MUSIC), or the statistical regularization method maximum-likelihood inspired adaptive robust iterative approach (MARIA), tackle the synthetic aperture radar (SAR) tomography (TomoSAR) inverse problem, achieving enhanced resolution, along with ambiguity levels reduction. Nevertheless, the aforementioned features come with a drawback, in order to suit the considered signal model, MUSIC and most regularization approaches require an appropriate setting of the involved parameters. In both cases, the right selection of these parameters is highly important, because the quality of the retrieved solutions depends on this. Consequently, this thesis treats several parameter selection criteria, adapted specifically to the TomoSAR scenario. On one hand, for regularization methods, we explore the Morozov’s discrepancy principle, the L-Curve, the Stein’s unbiased risk estimate and the generalized cross-validation to choose the regularization parameters. Otherwise, for the case of parametric methods like MUSIC, due to the differences on how they attack the problem, we consider a different methodology for the proper estimation of their parameters: the Kullback-Leibler information criterion to select the model order. After the incorporation of these criteria to MUSIC and MARIA, respectively, their capabilities are first analyzed through simulations, and later on, utilizing real data acquired from an urban area.