An inverse scattering reconstruction method for perfect electric conductor‐dielectric hybrid target based on physics‐inspired network

To improve the imaging accuracy in solving inverse scattering problems with mixed boundary conditions, the authors propose an inverse scattering imaging method based on a physics‐inspired neural network. First, the contrast source inversion and the T‐matrix method are unified to establish a combined...

Full description

Saved in:
Bibliographic Details
Published in:IET radar, sonar & navigation sonar & navigation, 2023-08, Vol.17 (8), p.1286-1298
Main Authors: Zhang, Qian‐qian, Yin, Cheng‐you, Li, An‐qi, Liu, Han
Format: Article
Language:English
Subjects:
computational electromagnetics
convolutional neural network
inverse scattering problem
microwave imaging
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:To improve the imaging accuracy in solving inverse scattering problems with mixed boundary conditions, the authors propose an inverse scattering imaging method based on a physics‐inspired neural network. First, the contrast source inversion and the T‐matrix method are unified to establish a combined parameter model, in which the characteristics of perfect electric conductor and dielectric scatterers are represented by the T‐matrix coefficients t and the dielectric contrast χdie ${\boldsymbol{\chi }}^{die}$ respectively. Considering the influence of singularity of conductor contrast, the PEC region and the DIE contrast are reconstructed separately. Then, an alternative parameter updating neural network method, called APU‐Net, is proposed to update the contrast of dielectric scatterers and the T‐matrix of perfect electric conductor scatterers alternately. The experimental results demonstrate the strong performance of APU‐Net for inverse scattering imaging of PEC‐DIE hybrid targets as well as its improved generalisation ability over existing methods. For a conductor‐dielectric hybrid target, APU‐Net separates the reconstruction of the PEC part from the DIE part, and realizes the alternate iterative updating of parameters, in which way to reduce the interference of conductors with the dielectric part in the process of reconstruction. Besides, physical information was added to the network training through numerical calculation modules, integrating seamlessly data and mathematical physics models.
ISSN:1751-8784
1751-8792
DOI:10.1049/rsn2.12419