Reconstruction of a Large and High-Contrast Penetrable Object by Using the Genetic and Levenberg-Marquardt Algorithms

Abstract

A large and high-contrast two-dimensional dielectric object is reconstructed from the measured scattered fields by using a hybrid optimization algorithm combining the genetic algorithm and the Levenberg-Marquardt algorithm. Ill-posedness is filtered out by using the effective angular modes excluding the evanescent modes, and multiple incidences of plane waves of multifrequencies are used to reconstruct a dielectric object of its size, 3 x 3 wavelengths, and its relative permittivity 4.0 in presence of 10% Gaussian noise in the scattered fields. Object size is enlarged by using the large discretized cells up to 0.5 wavelength divided by the square root of relative permittivity. (C) 1997 John Wiley & Sons, Inc.