Font Size: a A A

Sparse Reconstruction Methods In Electrical Resistance Tomography

Posted on:2016-02-23Degree:DoctorType:Dissertation
Country:ChinaCandidate:J ZhaoFull Text:PDF
GTID:1108330485955056Subject:Detection Technology and Automation
Abstract/Summary:PDF Full Text Request
As a new cross-section detection technique, Electrical resistance tomography(ERT) attracts more attentions due to its properties of non-introsive, non-radiative, low-cost and fast response. However, because of its illness and nonlinear, the spatial resolution is low, which leads to unfavorable reconstruction results. It has important significance for accurately recognize and monitor the real-time status of industrial processes. To reduce the influence of illness of inverse problem for electrical resistance tomography, this work employs the regularization methods to investigate it systematically. In order to improve the reconstruction accuracy, speed and noise robustness for ET, the following works are carried out:(1) For the selection of regularization parameter in L0 regularization method, a fast adaptive iterative threshold algorithm is proposed, which has good reconstruction results for edges, and the imaging speed is greatly improved.(2) The reconstruction accuracy, speed and noise robustness of L1 regularization method are investigated systematically. The advantages of L1 regularizaiton method in edge preservation and noise robustness are verified with small objects imaging; A hybrid regularization method is proposed to solve the low solving speed of L1 regularization method, which combined the projection method and sparse regularization method. The proposed method further improves the imaging speed and noise robustness of L1 regularization method.(3) In order to clearly investigate the different imaging effects between the L1 regularization method and L2 regularization method, an Lq-Lp( p,q≤≤≤≤2121) optimization framework with its solving method is proposed. Studies of simulation, experiments and analysis of norm theory with fixed q value and regularization parameter find that the reconstruction results of the optimization framework are improved then become inferior with the increase of the regularization index, and the noise robustness of the algorithm is declined with the increase of the index p in the regularization term.(4) The sparse method of sensitivity matrix and the projection method are employed to reduce the high demand of computer performance for three dimensional ERT with its large amount of data and illness problem. Furthermore, a dimensional reduction SIRT algorithm(DR-SIRT) that has high noise robustness and fast solving speed is proposed, which greatly lowers the calculation amount and memory usage.(5) As ERT can not be used to normally reconstruct two-phase stratified distribution with one non-conductive phase, the total variation(TV) algorithm is employed due to its sparse property. Based on the judgement method for the position of interface of two media, two image reconstruction methods including the valid data reconstruction method and the new sensitivity matrix reconstruction method are proposed. Simulation and experimental results verify the feasibility of the two reconstruction methods.
Keywords/Search Tags:Electrical Resistance Tomography, Regularization Methods, Sparse Reconstruction Algorithms, Inverse Problem, Iterative Threshold Algorithm, Projection Algorithm, Dimensional Reduction Method
PDF Full Text Request
Related items