Physics-Driven Neural Network for Solving Electromagnetic Inverse Scattering Problems

2025-07-22

AI Summary

To address the poor generalizability and heavy reliance on large-scale labeled datasets inherent in deep learning approaches for electromagnetic inverse scattering, this paper proposes a physics-driven neural network-based iterative reconstruction framework. The method embeds the electromagnetic wave propagation physics directly into the network architecture and jointly optimizes both network parameters and physical variables. A composite loss function is formulated by integrating measured scattered fields with physically informed prior constraints. Additionally, a subregion identification mechanism is introduced to enhance modeling accuracy for complex lossy scatterers. Crucially, the framework achieves high-robustness, rapid-convergence imaging without requiring extensive labeled training data. Numerical simulations and microwave anechoic chamber experiments validate its high reconstruction accuracy, strong stability across diverse scenarios, and broad spectral applicability—significantly extending the practical deployment boundary of deep learning in electromagnetic inverse scattering.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Computer Vision: Learning & Optimization for CV

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

In recent years, deep learning-based methods have been proposed for solving inverse scattering problems (ISPs), but most of them heavily rely on data and suffer from limited generalization capabilities. In this paper, a new solving scheme is proposed where the solution is iteratively updated following the updating of the physics-driven neural network (PDNN), the hyperparameters of which are optimized by minimizing the loss function which incorporates the constraints from the collected scattered fields and the prior information about scatterers. Unlike data-driven neural network solvers, PDNN is trained only requiring the input of collected scattered fields and the computation of scattered fields corresponding to predicted solutions, thus avoids the generalization problem. Moreover, to accelerate the imaging efficiency, the subregion enclosing the scatterers is identified. Numerical and experimental results demonstrate that the proposed scheme has high reconstruction accuracy and strong stability, even when dealing with composite lossy scatterers.

Problem

• Solving electromagnetic inverse scattering with limited generalization
• Optimizing physics-driven neural networks using scattered field constraints
• Improving imaging efficiency by identifying scatterer subregions

Innovation

• Physics-driven neural network optimizes hyperparameters iteratively
• Minimizes loss with scattered fields and prior constraints
• Identifies scatterer subregion to accelerate imaging