卷积神经网络在ECT图像重建上的应用

Application of Convolutional Neural Network in ECT Image Reconstruction

针对电容层析成像的图像重建精度较低的问题,提出一种基于卷积神经网络的ECT图像重建方法.首先阐述了ECT图像重建的基本原理,并利用COMSOL软件提取了大量的学习样本.然后以Landweber算法的图像重建结果作为初始状态,建立了卷积神经网络模型,并进行网络训练,保存训练完成的网络模型.最后选取样本以外的五种不同流型进行了仿真实验,实验结果表明,利用此算法所获取的重建图像相应指标要比LBP以及Landweber要好很多.所以该图像重建算法是一种有效且精度较高的图像重建算法.

In order to solve the problem of poor quality of image reconstruction in Electrical Capacitance Tomography(ECT),a image reconstruction method for the ECT was proposed based on the Convolutional Neural Networks.Firstly,the basic theory of ECT image reconstruction is expounded,a large number of learning samples were drawn by COMSOL software.Then,the convolution neural network model is established by using the image reconstruction result of Landweber algorithm,and the network model is trained and saved while the training is completed.Finally,the simulation experiments with five different flow regimes out of the training set were performed,the experiment results show that the corresponding index of reconstruction results by proposed method is much better than the LBP and Landweber.In conclution,the image reconstruction method is an efficient and highly accurate algorithm.