Image Reconstruction for Electrical Resistance Tomography Using Convolutional Neural Network and Visual Geometry Group

Electrical resistance tomography (ERT) is a process visualization technique. It has the advantages of non-radiation, non-invasiveness and fast response, and has been widely studied. Image reconstruction of ERT faces the challenges of nonlinearity and ill-posedness of inverse problem, which makes the traditional image reconstruction algorithms suffer from low resolution and poor reconstruction accuracy. To improve the image reconstruction accuracy of ERT, a novel deep learning method using convolutional neural network (CNN) and visual geometry group (VGG) is proposed. The network structure is based on CNN, and the convolution group adopts the convolution stacking method in the VGG model, and uses batch normalization and dropout strategies to improve network performance. Compared with the traditional image reconstruction methods and CNN network, simulation and experimental results show that the proposed CNN-VGG method has higher image reconstruction accuracy.