Research on ECT image reconstruction method based on Long Short-Term Memory Network (LSTM)

Abstract

In order to solve the problem of low image reconstruction accuracy of electrical capacitance tomography (ECT), this paper proposes an ECT image reconstruction method based on Long Short-Term Memory Network (LSTM). Based on the analysis and research of LSTM and ECT image reconstruction problems, the image reconstruction problem is transformed into a time series problem that LSTM can handle. Firstly, this paper proposes an image based encoding method of “rows” and “columns”, which organizes the images into N*M matrices according to “rows” and “columns” respectively. These N rows are considered as output “time series” with time steps. Then, according to the above method, the reconstructed images obtained using the Landweber method and setting images are encoded separately. Thus, the input and output sample libraries of the neural network are obtained. At the same time, the LSTM model was constructed and trained according to the “row” and “column” encoding methods. The above trained models are respectively used for inference, and the inference results are post-processed and fused, and the fusion results are used as the final ECT reconstruction image. Finally, simulation experiments were conducted. The experimental results have shown that the image error and correlation coefficient of the reconstructed image proposed in this paper are superior to Landweber algorithm, Tikhonov algorithm and convolutional neural network (CNN). This also provides a new approach and means for ECT image reconstruction algorithms.