Transverse Electric Inverse Scattering of Buried Conductors in a Slab Medium Using DSM and U-Net

Area of Science:

• Electromagnetics
• Computational Physics
• Inverse Problems

Background:

• Accurate reconstruction of buried conductive targets is crucial in various applications.
• Traditional methods like Method of Moments (MoM) and the Difference Equation Method (DEM) face challenges with complex media and noise.
• Electromagnetic inverse scattering problems require robust imaging techniques.

Purpose of the Study:

• To develop and validate a novel DSM-initialized U-Net framework for reconstructing the shape of perfectly conducting targets in a slab medium.
• To assess the performance of the proposed framework against standalone DEM under noisy conditions.
• To evaluate the impact of slab medium constraints on imaging accuracy.

Main Methods:

• Utilizing Transverse Electric (TE) waves for target irradiation and acquisition of scattered fields.
• Employing the Method of Moments (MoM) for forward scattering calculations.
• Implementing a Difference Equation Method (DEM) for preliminary shape reconstruction.
• Integrating DEM-derived images as input for a U-Net-based deep learning imaging model.

Main Results:

• The standalone DEM provided only approximate conductor profiles.
• The proposed DSM-U-Net framework achieved an average Normalized Root Mean Square Error (NRMSE) of 6.90% under 5% Gaussian noise.
• The framework demonstrated robustness, achieving 8.95% NRMSE when trained with 5% noise and tested with 10% noise.

Conclusions:

• The DSM-U-Net framework offers a significant advancement in electromagnetic inverse scattering for buried targets within slab media.
• The integration of deep learning with traditional methods enhances reconstruction accuracy and noise resilience.
• This approach provides a more reliable solution compared to standalone DEM, especially under challenging environmental conditions.