12th World Congress on Industrial Process Tomography
Spatial-wavelet Dual-domain Powered Complex-value Electrical Impedance Tomography (Cv-EIT)
Zichen, Wang1, Tao, Zhang1
1School of Electronic and Information Engineering, Tiangong University, Tianjin, China
Email: wangzc@tiangong.edu.cn, zhangtao@tiangong.edu.cn
ABSTRACT
Complex-value electrical impedance tomography (Cv-EIT) could reconstruct the conductivity and permittivity distributions simultaneously in the observation domain. Whereas, the inverse problem of Cv-EIT is ill-posed and nonlinear resulting the inaccurate admittivities and boundary distortions. In this letter, we propose a novel learning-based image reconstruction method with an ‘end-to-end’ manner to obtain high-quality Cv-EIT image. The two-stage framework is designed to implement the manifold transfer and image reconstruction. Specifically, the measured signals are transferred to the parameters distribution in the image domain using the iteration-based algorithm. Moreover, a spatial-wavelet dual-domain U-shaped network are proposed, where the low-frequency and high-frequency features are interacted using discrete wavelet transform and convolution/Transformer in the Encoder, bottleneck, and Decoder for conductivity and permittivity complementation. The experimental results show that our method has better performance in terms of quantitative metrics and visual results compared with competing methods. Meanwhile, the ablation study confirms the effectiveness of the spatial-wavelet feature fusion manner.
Keywords: Complex-value electrical impedance tomography, inverse problem, spatial information, wavelet features, convolution, Transformer.
Industrial Application: AI for tomography, Raw data processing such as direct inversion and high-speed reconstruction methods, PT in various applications such as medical imaging and NDT&E.
Sign-in to access the full text
Copyright © International Society for Industrial Process Tomography, 2025. All rights reserved.