Study On Image Reconstruction Algorithm Of Magnetic Induction Tomography

Magnetic Induction Tomography(MIT)is a new kind of the electrical magnetic characteristics tomography technology.Under the action of adscititious AC excitation magnetic field,the measured object produce eddy current because of the electromagnetic induction effect,when the measured object changes,intensity and distribution of eddy current will change accordingly.Then the inductive voltage is picked up through detection coils.Image reconstruction algorithm is used to reconstruct internal structure of the measured object.Magnetic Induction Tomography is safe,non-invasive,non-contact,cheap,etc.So it has a broad application prospect in the field of biomedical research,industrial process control,foreign body monitoring,geological exploration and other fields.MIT is still in the stage of laboratory research,how to apply the MIT to actual production as soon as possible is the objectives of researchers.On the base of reading mass home and abroad references,the theory and mathematical model of MIT have been analyzed and summarized.Some improved image reconstruction algorithm and measurement methods for MIT have been put forward in this paper.The main works of this paper are as follows:1?The math equations for MIT forward problem and image reconstruction are established based on Maxwell equations and principles of MIT.Using finite element software Comsol,a MIT system of 3-D simulation model is established,and the forward problem of MIT is calculated.Then the object impact on sensitive field and inductive voltage,which provided a valid instruction and data sources for the image reconstruction.2?Traditional Landweber algorithm was not high and convergence speed was slow,so thresholding Landweber iterative method is put forward.The method sets thresholding value on the basis of Landweber iterative method,adaptively regulates the thresholding parameter with the iteration,which can accelerate the convergence rate of the Landweber method.Finally,The experiment results have shown that the image reconstruction accuracy obtained by thresholding Landweber iterative method are better than traditional LBP and Landweber method,and the convergence speed is increased.3?Within the incomplete-data framework of MIT,the quality of image is low.So an image reconstruction method based on compressed sensing is put forward.The measurement data are compensated by the basic principle of the compressed sensing,then mathematical model is obtained.Finally,the experiment is executed,and image reconstruction error and correlation coefficient illustrated the effective.4?In order to improve the non-linear and ill-posed problem of MIT,the particle filter has the ability to solve nonlinear problems more effectively.A new image reconstruction method of MIT based on particle filtering algorithm is presented.Firstly,the principle of particle filtering is analyzed.Then the search process of the optimal solution for image reconstruction of MIT is described as system state estimation process,and the state space model is established.Finally,the simulation experiments with different flow regimes are performed,particle filtering is reflected the advantage.The results overcome the instability of image reconstruction,and obtained high quality image.5?On the basis of the simulation experiment for MIT,a complete set of measurement system has been designed for eddy current signal through the analysis of MIT technology and sensitivity research.Firstly,the overall structure of the system,the electrical characteristic parameters and the accuracy of the circuit are confirmed.Then,Hardware system includes excitation source,coils and phase detection device is designed.Then,the test experiments testing voltage value change rule is consistent with the law obtained in the previous studies,this verifies the stability and repeatability of this detection system.Finally,the preliminary image is reconstructed for aluminum bar,the experiment results show that traditional Landweber,compressed sensing and particle filtering algorithm are effective.This provides hint for further improving MIT system.