Yutong Du
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yutong Du has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Zicheng Liu
School of Electronics and Information, Northwestern Polytechnical University, Xian, China
5 shared publications - Yali Zong
School of Electronics and Information, Northwestern Polytechnical University, Xian, China
4 shared publications - Changyou Li
School of Electronics and Information, Northwestern Polytechnical University, Xian, China
3 shared publications Bazargul Matkerim
Department of Computer Science, Al-Farabi Kazakh National University, Almaty, Kazakhstan
2 shared publications- Bo Qi
Northwestern Polytechnical University
2 shared publications - Jingwei Kou
School of Electronics and Information, Northwestern Polytechnical University, Xian, China
2 shared publications
Latest Publications
- 1 July 2026
- Published by NDT.net GmbH & Co. KG in e-Journal of Nondestructive Testing
- Vol. 31 (7)
Physics-driven neural network solver for electromagnetic non-destructive evaluation
The non-destructive evaluation can be challenging due to the limited observation angles, sparse measurement points, and high computational burden [1]. Electromagnetic inverse scattering theories can be applied to improve the imaging performance. Artificial intelligence has been utilized to make use of data features, which can be learned based on a large dataset with a neural network. However, such data-driven methods suffer from limitations from data size and the generalization ability. A physics-driven neural network (PDNN) solver [2] was proposed. PDNN is purely physics-driven and the hyperparameters are updated only requiring the measured fields. However, its computational burden is still high and tens of seconds are required for the imaging which stops the application in nondestructive testing. To accelerate the convergence speed of the solver, we will introduce transfer learning [3] into PDNN. The pretrained initial model is obtained based on the sound pipeline structure, and the neural network weights are then updated based on the measured fields corresponding to materials with defects. The imaging examples of the pretrained and non-pretrained PDNN models are shown in Fig. 1, which demonstrate that the proposed imaging strategy enables faster convergence and higher efficiency (few seconds) with improved imaging accuracy.
- 1 July 2026
- Published by NDT.net GmbH & Co. KG in e-Journal of Nondestructive Testing
- Vol. 31 (7)
Physics-driven neural network solver for electromagnetic non-destructive evaluation
The non-destructive evaluation can be challenging due to the limited observation angles, sparse measurement points, and high computational burden [1]. Electromagnetic inverse scattering theories can be applied to improve the imaging performance. Artificial intelligence has been utilized to make use of data features, which can be learned based on a large dataset with a neural network. However, such data-driven methods suffer from limitations from data size and the generalization ability. A physics-driven neural network (PDNN) solver [2] was proposed. PDNN is purely physics-driven and the hyperparameters are updated only requiring the measured fields. However, its computational burden is still high and tens of seconds are required for the imaging which stops the application in nondestructive testing. To accelerate the convergence speed of the solver, we will introduce transfer learning [3] into PDNN. The pretrained initial model is obtained based on the sound pipeline structure, and the neural network weights are then updated based on the measured fields corresponding to materials with defects. The imaging examples of the pretrained and non-pretrained PDNN models are shown in Fig. 1, which demonstrate that the proposed imaging strategy enables faster convergence and higher efficiency (few seconds) with improved imaging accuracy.
- 1 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Antennas and Wireless Propagation Letters
- Vol. PP (15361225) , 1-5
A Fast Physics-Driven Fourier-Spectral Solver for Highly Nonlinear Inverse Scattering Problems
Physics-driven neural solvers can reconstruct strongly nonlinear inverse scattering scenes without supervised training, but their per-instance optimization over pixel-wise unknowns is computationally expensive. This letter presents a fast physics-driven Fourier-spectral solver for two-dimensional transverse-magnetic inverse scattering. The induced current is parameterized in a truncated Fourier basis, so that the reconstruction is carried out in a compact spectral space rather than on dense spatial grids. To improve robustness in high-contrast regimes, a contraction integral equation is incorporated into the physics-consistency model. In addition, a contrast-compensated operator is introduced to alleviate boundary attenuation caused by spectral truncation, and a bridge-suppressing loss is designed to reduce false connections between adjacent scatterers. A lightweight fully connected network, initialized from a back-propagation-based estimate, optimizes the spectral coefficients for each measurement instance. Numerical examples and Fresnel experimental data show that the proposed method preserves the main target topology and quantitative contrast with sub-second reconstruction time. Compared with representative spatial-domain physics-driven solvers, it achieves nearly two orders of magnitude acceleration while maintaining competitive reconstruction quality, which is promising for fast microwave imaging applications.
- 1 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Antennas and Wireless Propagation Letters
- Vol. PP (15361225) , 1-5
A Fast Physics-Driven Fourier-Spectral Solver for Highly Nonlinear Inverse Scattering Problems
Physics-driven neural solvers can reconstruct strongly nonlinear inverse scattering scenes without supervised training, but their per-instance optimization over pixel-wise unknowns is computationally expensive. This letter presents a fast physics-driven Fourier-spectral solver for two-dimensional transverse-magnetic inverse scattering. The induced current is parameterized in a truncated Fourier basis, so that the reconstruction is carried out in a compact spectral space rather than on dense spatial grids. To improve robustness in high-contrast regimes, a contraction integral equation is incorporated into the physics-consistency model. In addition, a contrast-compensated operator is introduced to alleviate boundary attenuation caused by spectral truncation, and a bridge-suppressing loss is designed to reduce false connections between adjacent scatterers. A lightweight fully connected network, initialized from a back-propagation-based estimate, optimizes the spectral coefficients for each measurement instance. Numerical examples and Fresnel experimental data show that the proposed method preserves the main target topology and quantitative contrast with sub-second reconstruction time. Compared with representative spatial-domain physics-driven solvers, it achieves nearly two orders of magnitude acceleration while maintaining competitive reconstruction quality, which is promising for fast microwave imaging applications.
- 10 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Antennas and Propagation
- Vol. 74 (7) , 6944-6953
Improved Physics-Driven Neural Network for Solving Inverse Scattering Problems
This paper presents an improved physics-driven neural network (IPDNN) framework for solving electromagnetic inverse scattering problems (ISPs). Unlike data-driven solvers that require large labeled datasets and often suffer from limited generalization, the proposed method performs instance-wise optimization under electromagnetic governing equations and measured scattered-field constraints. Three improvements over the original PDNN framework are introduced. First, a Gaussian-localized oscillation-suppressing window (GLOW) activation function is designed to stabilize convergence and enable a lightweight yet accurate network architecture. Second, a dynamic scatter subregion identification strategy is further developed to adaptively refine the computational domain, preventing missed detections and reducing computational cost. Third, transfer learning is incorporated to extend the solver’s applicability to practical scenarios, integrating the physical interpretability of iterative algorithms with the approximation capability of neural networks. Numerical and experimental results on representative Fresnel datasets show that the proposed IPDNN achieves up to a 57% reduction in reconstruction error while maintaining robust performance under noise levels up to 50%. Under the current implementation, the proposed FCN-based realization also requires substantially less training memory than the original PDNN. These results demonstrate that IPDNN provides a more accurate and practically extensible physics-driven framework for electromagnetic inverse scattering.
- 10 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Antennas and Propagation
- Vol. 74 (7) , 6944-6953
Improved Physics-Driven Neural Network for Solving Inverse Scattering Problems
This paper presents an improved physics-driven neural network (IPDNN) framework for solving electromagnetic inverse scattering problems (ISPs). Unlike data-driven solvers that require large labeled datasets and often suffer from limited generalization, the proposed method performs instance-wise optimization under electromagnetic governing equations and measured scattered-field constraints. Three improvements over the original PDNN framework are introduced. First, a Gaussian-localized oscillation-suppressing window (GLOW) activation function is designed to stabilize convergence and enable a lightweight yet accurate network architecture. Second, a dynamic scatter subregion identification strategy is further developed to adaptively refine the computational domain, preventing missed detections and reducing computational cost. Third, transfer learning is incorporated to extend the solver’s applicability to practical scenarios, integrating the physical interpretability of iterative algorithms with the approximation capability of neural networks. Numerical and experimental results on representative Fresnel datasets show that the proposed IPDNN achieves up to a 57% reduction in reconstruction error while maintaining robust performance under noise levels up to 50%. Under the current implementation, the proposed FCN-based realization also requires substantially less training memory than the original PDNN. These results demonstrate that IPDNN provides a more accurate and practically extensible physics-driven framework for electromagnetic inverse scattering.
- 5 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Antennas and Propagation
- Vol. 74 (2) , 1945-1956
Physics-Driven Neural Network for Solving Electromagnetic Inverse Scattering Problems
In recent years, deep learning-based methods have been introduced for solving inverse scattering problems (ISPs), but most of them heavily rely on large training datasets and suffer from limited generalization capability. In this paper, a new solving scheme is proposed where the solution is iteratively updated through a physics-driven neural network (PDNN), the hyperparameters of which are optimized by minimizing the loss function which incorporates the constraints from measured scattered fields and the prior information about scatterers. Unlike data-driven neural network solvers, PDNN does not require an offline training dataset. Its weights are iteratively updated based solely on the measured incident and scattered fields, similar in philosophy to conventional inverse algorithms but enhanced by neural-network flexibility. Thus, the generalization issue is eliminated. Moreover, to accelerate the imaging, a subregion enclosing the scatterers is automatically identified and used to restrict the computational domain. Numerical and experimental results demonstrate that the proposed scheme achieves high reconstruction accuracy and strong stability, even for complex and lossy scatterers.
- 5 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Antennas and Propagation
- Vol. 74 (2) , 1945-1956
Physics-Driven Neural Network for Solving Electromagnetic Inverse Scattering Problems
In recent years, deep learning-based methods have been introduced for solving inverse scattering problems (ISPs), but most of them heavily rely on large training datasets and suffer from limited generalization capability. In this paper, a new solving scheme is proposed where the solution is iteratively updated through a physics-driven neural network (PDNN), the hyperparameters of which are optimized by minimizing the loss function which incorporates the constraints from measured scattered fields and the prior information about scatterers. Unlike data-driven neural network solvers, PDNN does not require an offline training dataset. Its weights are iteratively updated based solely on the measured incident and scattered fields, similar in philosophy to conventional inverse algorithms but enhanced by neural-network flexibility. Thus, the generalization issue is eliminated. Moreover, to accelerate the imaging, a subregion enclosing the scatterers is automatically identified and used to restrict the computational domain. Numerical and experimental results demonstrate that the proposed scheme achieves high reconstruction accuracy and strong stability, even for complex and lossy scatterers.
- 12 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 63 (01962892) , 1-13
Quality-Factor-Inspired Deep Neural Network Solver for Solving Inverse Scattering Problems
Deep neural networks (DNNs) 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 concerned 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 analyses are performed to demonstrate the superiority of the proposed quality-factor-inspired DNN (QuaDNN) solver, and the imaging performance is finally verified by experimental imaging tests.
- 12 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 63 (01962892) , 1-13
Quality-Factor-Inspired Deep Neural Network Solver for Solving Inverse Scattering Problems
Deep neural networks (DNNs) 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 concerned 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 analyses are performed to demonstrate the superiority of the proposed quality-factor-inspired DNN (QuaDNN) solver, and the imaging performance is finally verified by experimental imaging tests.