A Model-Based Technique with ℓ_1 Minimization for Defect Detection and RCS Interpolation from Limited Data

摘要

Method of moments (MoM) codes have become increasingly capable and accurate for predicting the radiation and scattering from structures with dimensions up to several tens of wavelengths. In an earlier work, we presented a network model (NM) algorithm that uses a Gauss-Newton iterative nonlinear estimation method in conjunction with a CARLOS-3DTM MoM model to estimate the "as-built" materials parameters of a target from a set of backscatter measurements. In this paper, we demonstrate how the NM algorithm, combined with the basis pursuits (BP) ℓ_1 minimization technique, can be used to locate unknown defects (dents, cracks, etc.) on a target from a limited set of RCS pattern measurements. The advantage of ℓ_1 minimization techniques such as BP is that they are capable of finding sparse solutions to underdetermined problems. As such, they reduce the requirement for a priori information regarding the location of the defects and do not require Nyquist sampling of the input pattern measurements. We will also show how the BP solutions can be used to interpolate RCS pattern data that is undersampled or has gaps.