Research on ECT image reconstruction algorithm based on CycleGAN deep-learning network

At present, the development and application of electrical capacitance tomography (ECT) in the industrial field are becoming increasingly widespread. However, various algorithms still have a series of problems such as poor imaging quality, imaging blur, and excessive artifacts. There is an urgent need for clearer imaging models to deal with related problems. This article establishes a customized Cycle Adversarial Generative Neural Network (CycleGAN) model for electrical capacitance tomography. After generating a 2D model from the raw data, different manifold datasets are generated, and then the Cycle Adversarial Generative Neural Network is trained for deep learning. The evaluation results indicate that our ECT CycleGan model has advantages over traditional imaging algorithms in clearer ECT images, providing reference value for further optimization and improvement of capacitance layer analysis imaging algorithms.