Research On Sensitivity Matrix Calculation And Image Reconstruction Algorithm For Electrical Capacitance Tomography

Electrical Capacitance Tomography(ECT)is a method for measuring the spatial permittivity distribution information in a tube or container.Because of its non-intrusive sensing,fast response,easy portability,and low cost,it has been used in many other fields.At present,the difficulties in the ECT imaging process mainly focus on the design of image reconstruction algorithms and the selection and use under different environmental conditions.In view of this kind of problem,this article mainly studies from the following aspects.Firstly,for the ECT imaging system based on 12 electrodes,the system structure,mathematical principle and reconstruction model of ECT imaging are expounded,and the forward and inverse problems of ECT are deeply analyzed.The method of data normalization processing in ECT is studied,the principle analysis and the advantages and disadvantages of common image reconstruction algorithms are analyzed,and the experimental simulation environment is set up to collect experimental data.Secondly,the problems of image blur and low efficiency caused by non-linearization of standard sensitivity matrix are addressed.This paper proposes a K-means clustering algorithm based on filtering technology.The K-means clustering algorithm is used to improve the standard sensitivity matrix.At the same time,the threshold filtering technology is used to limit the range of sensitivity in each projection direction in the pipeline and improve the relative sensitivity in low-sensitivity areas.The experiment verifies the effectiveness of the algorithm and concludes that the algorithm is suitable for real-time fast imaging and clear judgment of the medium distribution in the pipeline.Finally,for complex situations,the lack of prior information in the inverseproblem caused by regularization leads to the problem of low imaging accuracy.A regularized image reconstruction algorithm based on extreme learning machines is proposed.Improvements are made on the basis of the extreme learning machine principle,and the prior information of the target domain is obtained through the improved extreme learning machine prediction model.A new cost function is constructed to encapsulate the prior information,and a spatial regularizer and a temporal regularizer are introduced to enhance the prior sparseness.A new numerical method composed of the Bregman splitting algorithm and FIST technology is used to solve the cost function to obtain the final imaging result.Simulation experiments verify the effectiveness of the algorithm and conclude that the algorithm is suitable for complex scenes that require high-precision restoration of the distribution of the multiphase medium in the pipeline and do not require real-time imaging.