A CNN-based Image Reconstruction Scheme for Complex-valued Multi-frequency Electrical Capacitance Tomography

Complex-valued multi-frequency electrical capacitance tomography (CVMF-ECT) is a newly developed imaging method for both industrial and medical applications, such as non-invasive fault detection and multi-phase flow measurement. It can simultaneously reconstruct spectra conductivity and permittivity properties of testing object within the region of interest (ROI), which can reveal more characteristics of conductive samples compared to the traditional single-frequency ECT. However, currently CVMFECT is still in small-scale trials, since the traditional iterative algorithms used by CVMF-ECT to reconstruct images have numerous iteration times and slow imaging speeds, result in poor imaging qualities. In this paper, an image reconstruction network namely SegNet is employed to effectively utilize the relationships between multi-frequency spectra data and the real distributions of dielectrics. Numerical simulations validate the superior performance of SegNet compared with traditional imaging reconstruction approaches such as Local Binary Patterns and Tikhonov Regularization.