| Tomography has attracted much attention due to its non-invasive and non-disturbing properties.Among them,electrical tomography and ultrasonic tomography have good application prospects in the industrial and medical fields due to their advantages such as: low cost,non-radiation,wide measurement range.However,in the image reconstruction process of electrical and ultrasonic tomography,due to the inherent non-linearity and ill-posedness of the sensitive principle,the reconstructed image has low spatial resolution and poor real-time performance,which is difficult to meet the needs of practical applications.Therefore,improving the accuracy and real-time performance of image reconstruction by electrical and ultrasonic tomography methods is of great significance for its application in industrial production processes and medical monitoring.The research work focuses on electrical resistance tomography(ERT)developed in electrical modal research and ultrasonic transmission tomography(UTT)widely used in ultrasonic modal research.For ERT,due to the low spatial resolution and poor real-time performance of image reconstruction;UTT has obvious sparsity,commonly used image reconstruction results have serious artifacts and large shape errors;since different modality has different physical background,ERT and UTT dual-modality image reconstruction is difficult to achieve effective fusion and affect the quality of the reconstructed image.Based on the thorough discussion of the effective information obtained by ERT and UTT,the image reconstruction method based on the compressed sensing theory is in-depth studied.Specific research work includes:(1)Based on the summary and analysis of the research process of tomography,especially ERT,UTT and dual-modality tomography,and the analysis and refinement of the characteristics of compressed sensing theory,in order to solve the problem of electrical tomography,which is represented by ERT,and ultrasonic tomography which is represented UTT,that they suffered low spatial resolution and poor real-time performance.According to the different sensitivity characteristics of different sensing field to different medium and the complementary of the two sensitive fields.An image reconstruction method based on the compressed sensing theory is proposed,which can effectively improve the test sensitivity range of ERT and UTT and improve the imaging quality.The electrical and ultrasonic dual modalities are effectively integrated,which realizes ERT and UTT dual-modality image reconstruction.(2)Based on the linear optimization method of compressed sensing theory,in order to solve the ill-posed problem in electrical resistance tomography,based on the analysis of the characteristics of the orthogonal matching pursuit(OMP)algorithm and the factors affecting the image reconstruction,the modified OMP(MOMP)algorithm is proposed.By adding the self-adaptive of number of iteration and the continuity constraints of the solution set to the MOMP algorithm,it is suitable for solving the ERT inverse problem.To compare the image reconstruction result by MOMP algorithm with image reconstruction result by the non-iterative algorithm and iterative algorithm,the image reconstruction quality is improved,and the MOMP image reconstruction speed is faster than iterative algorithm.Based on the implementation of MOMP algorithm,in order to improve the real-time performance of ERT image reconstruction,we further propose ERT compression sampling strategy,which can greatly improve the real-time performance of the image at the expense of a small amount of imaging accuracy.(3)Based on the nonlinear optimization method of compressed sensing,in order to solve the nonlinear characteristic of ERT image reconstruction process,by summing up the characteristics of ERT image reconstruction process into the objective function of the optimization method,the ERT image reconstruction problem is converted into a nonlinear multi-objective optimization problem.The nonlinear multi-objective optimization problem is solved by homotopy method to improve the convergence and reduce the sensitivity of the convergence process to the initial value.Test experiments based on different models show that the nonlinear compressed sensing method can effectively solve the ERT inverse problem.The convergence of the solution process is good,and the convergence process is not sensitive to the initial value.The image quality parameters of non-linear compressed sensing image reconstruction algorithm are significantly improved compared with those of ERT reconstruction algorithm.(4)Based on the sparsity method of compressed sensing theory,SP-OMP(sparse OMP)algorithm is proposed to solve the problems of low resolution,large star-shape shadow and large shape error of image reconstruction results of LBP algorithm and ART algorithm in UTT.SP-OMP algorithm makes use of the sparse characteristics of UTT measurement data and adds sparse constraints and continuity constraints of the solution set on the basis of orthogonal matching pursuit algorithm,making it more suitable for solving the inverse problem of UTT.Experimental results based on different models show that SP-OMP algorithm improves the quality of image reconstruction and the accuracy of shape reconstruction compared with other frequently used UTT image reconstruction algorithms;SP-OMP algorithm has the highest image quality compared with other frequently used modifications of OMP algorithms.(5)Based on the research results of single modality compressed sensing image reconstruction method,in order to solve the problem that the measurement information obtained from different physical fields is hard to be deeply fused in the imaging process due to different physical dimension of different modalities,based on the derivation of the unified mathematical model of dual-modality tomography,a dual-modality tomography method,electrical ultrasonic projection sorting algorithm(EUPS)is proposed for information fusion between different modalities.Since the EUPS algorithm avoid the error superimposition in the dual-modality imaging process.The simulation and experimental results show that,comparing with single modality image algorithms,EUPS method has better imaging accuracy and anti-noise performance.Dynamic experiments on the bubbly tower show that the imaging speed of EUPS method can meet the speed requirement to monitor the bubble rising process in vertical gas-liquid two-phase flow and satisfy the online imaging requirement of industrial processes. |
PDF Full Text Request