International Society for Industrial Process Tomography

Deep Learning Based Image Reconstruction for Electrical Capacitance Tomography

Jin Zheng, Lihui Peng^*

Department of Automation, Tsinghua University, Beijing, China

^*Email: lihuipeng@mail.tsinghua.edu.cn

ABSTRACT

Research on image reconstruction of electrical capacitance tomography (ECT) has developed decades and made great achievements. However, there is still a need to find new algorithm framework to make ECT image reconstruction faster and better. Recent years, deep learning theory, which is based on different artificial neural networks and good at mapping complicated nonlinear functions, is flourishing and adopted in many fields. In this paper, a supervised deep autoencoder neural network, including an encoder and a decoder, is proposed to solve both the forward problem and the inverse problem of ECT, and an iteration method based on the proposed deep autoencoder is used to promote the image reconstruction quality. A large-scale dataset consisting of 40000 pairs of instances, of which each pair of sample has a capacitance vector and corresponding permittivity distribution vector, is used to train and test the performance of the deep autoencoder by 10-fold cross validation. Furthermore, data regarding flow patterns not in training dataset are used to test the generalization ability of the network and the effect of the iteration method. Reconstructed images and quantitative criteria show that the deep autoencoder and the iteration method based on it provide improved results.

Keywords electrical capacitance tomography, image reconstruction, deep learning, autoencoder, iteration

Industrial Application General