Quantitative Electromagnetic Inversion of Irregular Scatterers Based on a Threefold Hybrid Method

Abstract

This article presents a threefold hybrid electromagnetic inversion method that combines the linear sampling method (LSM), the convolutional neural network (CNN), and the full-wave inversion (FWI). The whole inversion procedure includes three steps. First, the LSM qualitatively reconstructs the preliminary shapes and locations of the scatterers. Then, the CNN U-Net is used to further refine the shapes of the scatterers. At last, the Born iterative method (BIM) is implemented to quantitatively invert for the permittivity and conductivity of inhomogeneous scatterers or multiple homogeneous scatterers inside the downsized inversion domain. Numerical experiments show that compared with the pure FWI method BIM, the proposed hybrid method can achieve both higher reconstruction accuracy and lower computational cost. Besides this superiority, the proposed hybrid method also has strong adaptability to multiple scatterers with high contrasts even when the measured field data are contaminated by large noise. Laboratory experimental data are also used to verify the feasibility of the proposed method.