Research On ECT Image Reconstruction Algorithm Based On Convolutional Neural Network

Electrical Capacitance Tomography Technology(ECT)which was produced in the 1980 s,is mainly used to research the distribution of materials with different permittivity in a pipeline or area.ECT technology because of its many advantages,such as simple structure,non-contact,low cost,wide range of detection,and so on,has a good prospect in the future of various industrial fields,such as petroleum,chemical industry,etc.After many years of research and development,ECT technology has made great progress,but there is still existence of soft field effect problems,which cannot be avoided and completely solved,thus,the image reconstruction accuracy of ECT system is not very satisfactory.So,it is very important to research the reconstruction algorithm of electrical capacitance tomography system in depth.In the future,some scholars still need to study and research the technology deeply.According to a great quantity of related literature at home and abroad,the background and significance of ECT technology,recently research on home and abroad and application area of ECT system are sketched in this paper.The basic composition and mathematical theory of the ECT system is elaborated in detail.The basic structure and working principle of convolutional neural network(CNN)are introduced in detail,and aiming at the low accuracy of ECT reconstruction image,a comprehensive application method of ECT image reconstruction based on CNN network is proposed,the following work is accomplished:Firstly,in view of the fact that the fixed sensitivity matrix provides less information in ECT image reconstruction at present,a method of updating the sensitivity matrix of ECT system reconstruction image based on CNN network is proposed.The method,based on lots of sensitive field samples extracted from the COMSOL software,sets up the network model,and then the model,which takes the value of the Landweber algorithm as the input,is used to update the sensitivity matrix and used in ECT image reconstruction.Finally,the simulation of the models isprovided,and the reconstructed image using the fixed sensitivity matrix is compared,the results show that the updating of the sensitivity matrix could improve the image quality,which is proven to be efficiency.Secondly,in view of the poor quality of image reconstruction in ECT system nowadays,this paper proposes a method of solving ECT image reconstruction problem using CNN network.The method,also based on lots of flow pattern samples extracted,sets up the network model.Then,the dielectric constant distribution of several typical flow patterns is reconstructed based on the network model,which takes the value of the Landweber algorithm as the input.Finally,the simulation of the models is provided,and the reconstructed image using other algorithms is compared,the results show that the method is proven to be efficiency.Finally,in order to further improve the accuracy of image reconstruction in ECT system,the sensitivity field updating method based on CNN network is combined with the method of solving ECT image reconstruction problem using CNN network,and a method of ECT image reconstruction based on CNN network is proposed in this paper,firstly,two network models are obtained according to the previous method,then the image reconstruction results of the Landweber algorithm are used in the network model of the sensitivity matrix to obtain the updated sensitivity matrix,and the flow pattern network model takes the reconstructed image of Landweber algorithm as the input,and then the field distribution is reconstructed.Finally,a simulation experiment is carried out to verify the proposed method.The test results show that the algorithm proposed in this paper,can effectively reflect the field situation of the reconstruction area and its relative error and correlation coefficient are all superior to LBP and Landweber algorithms,and it also provides an effective method and means for the ECT image reconstruction algorithm.