Reconstruction of Subsurface Objects by LSM and FWI From Limited-Aperture Electromagnetic Data

Abstract

This article presents a hybrid 3-D electromagnetic (EM) full-wave inversion (FWI) method for the reconstruction of subsurface objects illuminated by an antenna array with the limited aperture. The 3-D linear sampling method (LSM) is first used to qualitatively reconstruct the rough shapes and locations of the subsurface objects. Then, the 3-D convolutional neural network (CNN) U-Net is used to further refine the images of the unknown objects. Finally, the Born iterative method (BIM) is implemented to quantitatively invert for the dielectric parameters of subsurface inhomogeneous objects or multiple homogeneous objects in the restricted image regions. Numerical simulations show that, compared with the pure FWI method BIM, the proposed hybrid method can reconstruct subsurface 3-D objects from limited-aperture EM data with both higher accuracy and lower computational cost. In addition, the proposed hybrid method also shows a strong antinoise ability for the reconstruction of multiple subsurface objects.