Research On ECT Image Reconstruction Based On Memetic Algorithm

Electrical capacitance tomograhy(ECT)technology is a relatively mature process tomography technology with good development potential.It realizes the image reconstruction of the dielectric distribution in the region according to the capacitance change caused by the change of dielectric constant in the measured area.This technology has gradually become a research hotspot in the field of industrial non-destructive testing due to the various advantages,such as non-contact,simple structure,low cost,safety and radiation free,and wide application range.ECT image reconstruction is a serious nonlinear,pathological,and indefinite inverse problem.Since the reconstruction quality of images directly determines the application and development of ECT technology,improving the quality of imaging has always been a key and difficult point in the research of ECT technology.This thesis aims to improve the quality of reconstructed images by studying the improvement of the ECT forward problem model and the algorithm for solving the inverse problem.The main work achievements are as follows:(1)The advantages and disadvantages of several commonly used ECT imaging algorithms are analyzed in detail,and mathematical calculation methods are given.Finite element simulation models of various typical media distributions were established by COMSOL Multiphysics.(2)An improved ECT forward problem model is proposed.First,the sensitivity matrix is smoothed and preprocessed by the Butterworth low-pass filter to filter out the peaks and burrs that are likely to cause artifacts,and to alleviate the inhomogeneity of the sensitivity matrix;Redundant information in high-dimensional data reduces the condition number of the sensitivity matrix and alleviates the ill-conditioned nature of the ECT inverse problem.Then,the commonly used Landweber algorithm and SIRT algorithm are used to solve the inverse problem,which verifies the effectiveness of improving the forward problem model.(3)A new solution method is proposed for the inverse problem of ECT.Firstly,the objective function with total variation as the regular term is constructed,which better maintains the boundary information of the image.Then,the population-based global search and individualbased local search are combined by memetic algorithm to improve the probability of obtaining the global optimal solution.Combining the improved forward problem model,the new algorithm significantly improves the quality of reconstructed images.The effectiveness of the proposed method was verified through simulation experiments by Landweber and SIRT as comparative algorithms.(4)An experimental system was established to collect capacitance data corresponding to empty field,full field and four typical media distributions.Perform data analysis and image reconstruction on the collected capacitance values,and the proposed algorithm is effective and feasible by experimental verification.