Duvar arkası görüntüleme için derin öğrenme algoritmaları

Abstract

Ters saçılmaya dayalı hızlı duvar arkası görüntüleme (DAG), birçok güvenlik ve sivil uygulama için önemli bir ihtiyaçtır. Ancak ters saçılma tabanlı DAG, çözümü yinelemeli olarak elde edilen kötü koşullu ve doğrusal olmayan denklem takımlarının çözülmesini gerektirir; örneğin bu amaçla Distorted Born iteratif yöntemi kullanılabilir. Bu yinelemeli çözüm süreçleri uzun hesaplama süreleri gerektirmekte ve çoğu zaman yakınsamama problemiyle karşılaşmakta, dolayısıyla ters saçılmanın gerçek dünya DAG senaryolarındaki uygulanabilirliğini sınırlamaktadır. Bu sorunları ele almak amacıyla bu çalışmada, fiziğe dayalı modellemeyi sinir ağlarının etkin öğrenme kabiliyetiyle birleştiren derin öğrenme destekli ters saçılma algoritmaları incelenmektedir. Birkaç Distorted Born iterasyonu sonrasında elde edilen görüntüler girdi olarak kullanılarak, duvar arkasındaki saçıcıların nihai yüksek kaliteli görüntülerini elde etmek için üç farklı sinir ağı mimarisi—geleneksel evrişimsel sinir ağı (CNN), U-net ve maske bölge tabanlı CNN (Mask R-CNN)—kullanılmış ve performansları karşılaştırılmıştır. Bu üç teknik arasında geleneksel CNN saçıcı konumlarını kestirmekte ve dielektrik profilleri geri kazandırmakta, U-net saçıcıları bölütlemekte, Mask R-CNN ise saçıcıları tespit etmektedir. Sayısal sonuçlar, U-net’in en yüksek doğruluğu sağladığını; Mask R-CNN’in ise rekabetçi doğruluk sunarken sınır kusurlarını giderdiğini ve bu yönüyle DAG uygulamaları için oldukça etkili bir yöntem olduğunu göstermektedir.

Keywords

• Evrişimsel sinir ağı (CNN)
• derin öğrenme
• ters saçılma
• maske bölge tabanlı CNN (Mask R-CNN)
• duvar arkası görüntüleme (DAG)
• U-Net

Project Number

Nanyang Technological University, Start-Up Grant, Award No. 001096-00001

Deep learning algorithms for through-the-wall imaging

Abstract

Fast through-the-wall imaging (TWI) by inverse scattering is highly desirable for many security and civilian applications. TWI by inverse scattering requires the solution of an ill-posed and nonlinear set of equations whose solution is attained iteratively (e.g., via a distorted Born iterative method). The iterative solution requires long execution times and often does not converge, limiting the applicability of inverse scattering to real-world TWI scenarios. To address these issues, this study investigates deep learning-augmented inverse-scattering schemes that combine physics-based modeling with the learning capability of neural networks. Using images after a few distorted Born iterations as input, three neural networks, namely a traditional convolutional neural network (CNN), U-net, and mask region-based CNN (Mask R-CNN) are leveraged to obtain final high-quality images of the scatterers behind the walls and their performance is compared. Among these three techniques, the traditional CNN regresses scatterer positions and restores dielectric profiles, while U-net segments scatterers and Mask R-CNN detects scatterers. Numerical results show that U-net achieves the highest accuracy, while Mask R-CNN offers competitive accuracy and removes boundary defects, making it highly effective for TWI applications.

Keywords

• Convolutional neural network (CNN)
• deep learning
• inverse scattering
• Mask R-CNN
• through-the-wall imaging
• U-net

Supporting Institution

Nanyang Technological University

Project Number

Nanyang Technological University, Start-Up Grant, Award No. 001096-00001

Ethical Statement

N/A

Thanks

This work was conducted during the PhD studies of Xiaofan Jia under the supervision of Prof. Abdulkadir C. Yucel (the main/primary author). The work was supported by Nanyang Technological University via a Start-Up Grant (Award No. 001096-00001) awarded to Prof Yucel. The authors are thankful to Dr. Luis J. Gomez for fruitful discussions during the implementation of the distorted Born iterative method.

References

1. Chu, Y., Xu, K., Zhong, Y., Ye, X., Zhou, T., Chen, X., and Wang, G., Fast microwave through wall imaging method with inhomogeneous background based on Levenberg–Marquardt algorithm. IEEE Transactions on Microwave Theory and Techniques, 67, 1138–1147, (2019).
2. Chen, R., Wei, Z., and Chen, X., Three dimensional through-wall imaging: Inverse scattering problems with an inhomogeneous background medium. Proceedings, 2015 IEEE 4th Asia-Pacific Conference on Antennas and Propagation (APCAP), 505–506, (2015).
3. Chen, X., Wei, Z., Li, M., and Rocca, P., A review of deep learning approaches for inverse scattering problems. Progress In Electromagnetics Research, 167, 67–81, (2020).
4. Wei, Z., and Chen, X., Deep-learning schemes for full-wave nonlinear inverse scattering problems. IEEE Transactions on Geoscience and Remote Sensing, 57, 1849–1860, (2019).
5. Slaney, M., Kak, A., and Larsen, L., Limitations of imaging with first-order diffraction tomography. IEEE Transactions on Microwave Theory and Techniques, 32, 860–874, (1984).
6. Xiao, J., Li, J., Chen, Y., Han, F., and Liu, Q. H., Fast electromagnetic inversion of inhomogeneous scatterers embedded in layered media by Born approximation and 3-D U-Net. IEEE Geoscience and Remote Sensing Letters, (2019).
7. Wang, Y., and Chew, W. C., An iterative solution of the two-dimensional electromagnetic inverse scattering problem. International Journal of Imaging Systems and Technology, 1, 100–108, (1989).
8. Chew, W. C., and Wang, Y., Reconstruction of two-dimensional permittivity distribution using the distorted Born iterative method. IEEE Transactions on Medical Imaging, 9, 218–225, (1990).

9. Van Den Berg, P. M., and Kleinman, R. E., A contrast source inversion method. Inverse Problems, 13, 1607, (1997).
10. Yao, Y. M., Sha, W. E. I., and Jiang, L., Two-step enhanced deep learning approach for electromagnetic inverse scattering problems. IEEE Antennas and Wireless Propagation Letters, 18, 2254–2258, (2019).
11. Wei, Z., and Chen, X., Physics-inspired convolutional neural network for solving full-wave inverse scattering problems. IEEE Transactions on Antennas and Propagation, 67, 6138–6148, (2019).
12. Sun, J., Niu, Z., Innanen, K. A., Li, J., and Trad, D. O., A theory-guided deep-learning formulation and optimization of seismic waveform inversion. Geophysics, 85, R87–R99, (2020).
13. Guo, R., Song, X., Li, M., Yang, F., Xu, S., and Abubakar, A., Supervised descent learning technique for 2-D microwave imaging. IEEE Transactions on Antennas and Propagation, 67, 3550–3554, (2019).
14. Chen, G., Shah, P., Stang, J., and Moghaddam, M., Learning-assisted multi-modality dielectric imaging. IEEE Transactions on Antennas and Propagation, 68, 2356–2369, (2019).
15. Sanghvi, Y., Kalepu, Y., and Khankhoje, U. K., Embedding deep learning in inverse scattering problems. IEEE Transactions on Computational Imaging, 6, 46–56, (2020).
16. Sun, Y., Xia, Z., and Kamilov, U. S., Efficient and accurate inversion of multiple scattering with deep learning. Optics Express, 26, 14678–14688, (2018).
17. Li, L., Wang, L. G., Teixeira, F. L., Liu, C., Nehorai, A., and Cui, T. J., DeepNIS: Deep neural network for nonlinear electromagnetic inverse scattering. IEEE Transactions on Antennas and Propagation, 67, 1819–1825, (2019).
18. Khoshdel, V., Ashraf, A., and LoVetri, J., Enhancement of multimodal microwave-ultrasound breast imaging using a deep-learning technique. Sensors, 19, 4050, (2019).
19. Ye, X., Bai, Y., Song, R., Xu, K., and An, J., An inhomogeneous background imaging method based on generative adversarial network. IEEE Transactions on Microwave Theory and Techniques, (2020).
20. Xu, K., Qian, Z., Zhong, Y., Su, J., Gao, H., and Li, W., Learning-assisted inversion for solving nonlinear inverse scattering problem. IEEE Transactions on Microwave Theory and Techniques, 71, 2384–2395, (2022).
21. Hu, H.-J., Xiao, L.-Y., Li, J., and Liu, Q. H., A hybrid forward-inverse neural network with the transceiver-configuration-independent technique for the wideband electromagnetic inverse scattering problem. IEEE Transactions on Geoscience and Remote Sensing, 62, 1–11, (2024).
22. Liu, Z., Roy, M., Prasad, D. K., and Agarwal, K., Physics-guided loss functions improve deep learning performance in inverse scattering. IEEE Transactions on Computational Imaging, 8, 236–245, (2022).
23. Zhou, H., Cheng, Y., Zheng, H., Liu, Q., and Wang, Y., Deep unfolding contrast source inversion for strong scatterers via generative adversarial mechanism. IEEE Transactions on Microwave Theory and Techniques, 70, 4966–4979, (2022).
24. Yin, T., Wang, C.-F., Xu, K., Zhou, Y., Zhong, Y., and Chen, X., Electric flux density learning method for solving three-dimensional electromagnetic scattering problems. IEEE Transactions on Antennas and Propagation, (2022).
25. Guo, R., Lin, Z., Shan, T., Song, X., Li, M., Yang, F., Xu, S., and Abubakar, A., Physics embedded deep neural network for solving full-wave inverse scattering problems. IEEE Transactions on Antennas and Propagation, 70, 6148–6159, (2021).
26. Yao, H. M., Guo, R., Li, M., Jiang, L., and Ng, M. K. P., Enhanced supervised descent learning technique for electromagnetic inverse scattering problems by the deep convolutional neural networks. IEEE Transactions on Antennas and Propagation, 70, 6195–6206, (2022).
27. Zhou, H., Xu, C., Cheng, Y., Liu, Q., Hu, S., and Wang, Y., Self-supervised learning of physics-guided deep unfolding parallel networks for solving nonlinear inverse scattering problems. IEEE Transactions on Microwave Theory and Techniques, 72, 5204–5217, (2024).
28. Huang, Q., Li, C., Ye, X., Xu, K., and Song, R., Meta-learning-assisted untrained neural network for electromagnetic inverse scattering problems. IEEE Transactions on Antennas and Propagation, (2025).
29. Song, R., Li, M., Xu, K., Li, C., and Chen, X., Electromagnetic inverse scattering with an untrained SOM-net. IEEE Transactions on Microwave Theory and Techniques, 70, 4980–4990, (2022).
30. Hu, Y.-D., Wang, X.-H., Zhou, H., and Wang, L., A priori knowledge-based physics-informed neural networks for electromagnetic inverse scattering. IEEE Transactions on Geoscience and Remote Sensing, 62, 1–9, (2024).
31. LeCun, Y., Bengio, Y., and Hinton, G., Deep learning. Nature, 521, 436–444, (2015).
32. Ronneberger, O., Fischer, P., and Brox, T., U-Net: Convolutional networks for biomedical image segmentation. Proceedings, International Conference on Medical Image Computing and Computer-Assisted Intervention, 234–241, Springer International Publishing, Cham, (2015).
33. He, K., Gkioxari, G., Dollár, P., and Girshick, R., Mask R-CNN. Proceedings, IEEE International Conference on Computer Vision, 2961–2969, (2017).
34. Paszke, A., Gross, S., Chintala, S., Chanan, G., Yang, E., DeVito, Z., Lin, Z., Desmaison, A., Antiga, L., and Lerer, A., Automatic differentiation in PyTorch. Proceedings, 31st Conference on Neural Information Processing Systems, (2017).
35. Kingma, D. P., and Ba, J., Adam: A method for stochastic optimization. arXiv Preprint arXiv:1412.6980, (2014).
36. Jadon, S., A survey of loss functions for semantic segmentation. arXiv Preprint arXiv:2006.14822, (2020).
37. Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L., ImageNet: A large-scale hierarchical image database. Proceedings, 2009 IEEE Conference on Computer Vision and Pattern Recognition, 248–255, (2009).
38. Xie, S., Girshick, R., Dollár, P., Tu, Z., and He, K., Aggregated residual transformations for deep neural networks. Proceedings, IEEE Conference on Computer Vision and Pattern Recognition, 1492–1500, (2017).
39. Wu, Y., Kirillov, A., Massa, F., Lo, W., and Girshick, R., Detectron2. GitHub, https://github.com/facebookresearch/detectron2, (2019).