Quality-factor inspired deep neural network solver for solving inverse scattering problems

TL;DR

This work addresses electromagnetic inverse scattering problems by introducing a quality-factor guided training paradigm (QuaDNN). It combines a ReSE-U-Net architecture with residual and SE attention and a physics-informed loss that blends data fitting, near-field constraints, and smoothness priors. The quality factor $Q_{ ext{BP}}$ guides dataset composition to emphasize informative, hard-to-predict samples, and the loss hyperparameter $eta$ is tuned to balance constraints and reconstruction fidelity. Across extensive numerical tests and experimental data, QuaDNN with mixed loss ($L^{ ext{Mix}}$) demonstrates superior imaging performance, robustness to SNR variations, and strong generalization to complex and overlapping scatterers, indicating practical potential for ISP imaging tasks.

Abstract

Deep neural networks have been applied to address electromagnetic inverse scattering problems (ISPs) and shown superior imaging performances, which can be affected by the training dataset, the network architecture and the applied loss function. Here, the quality of data samples is cared and valued by the defined quality factor. Based on the quality factor, the composition of the training dataset is optimized. The network architecture is integrated with the residual connections and channel attention mechanism to improve feature extraction. A loss function that incorporates data-fitting error, physical-information constraints and the desired feature of the solution is designed and analyzed to suppress the background artifacts and improve the reconstruction accuracy. Various numerical analysis are performed to demonstrate the superiority of the proposed quality-factor inspired deep neural network (QuaDNN) solver and the imaging performance is finally verified by experimental imaging test.

Figures (13)

• Figure 1: The imaging scheme of the concerned two-dimensional inverse scattering problem.
• Figure 2: Distribution of samples according to quality factor and relative permittivity value.
• Figure 3: Composition of training dataset $T_{\text{QBP}}$ according to quality factor $Q_{\text{BP}}$.
• Figure 4: ReSE-U-Net architecture used to solve ISPs in this paper. The input consists of two channels representing the real and imaginary parts of BP imaging result.The architecture integrates residual connection, channel attention mechanism and feature transformation layer into U-Net.
• Figure 5: Image reconstruction of digit-like (left part) and polygon-like (right part) profiles using U-Net and ReSE-U-Net architecture when measured scattered fields are corrupted by Gaussian noises with SNR = 5dB.