The Research Of Microwave Tomography Algorithm Based On Improved Newton Method

Microwave tomography(MWT)is a new industrial tomography detection technology developed in recent years.It has the characteristics of non-contact and safety,and is widely used in various fields.With the development of science and technology,people have put forward higher requirements for the accuracy and speed of detection.In the actual environment,the overall performance of the MWT system cannot be directly improved from the hardware aspect due to the equipment cost,signal interference between microwave sensors and the characteristics of the substance to be measured.Therefore,it is of great significance to improve the accuracy and speed of the MWT system by studying reconstruction algorithms without adding equipment.The thesis first elaborated on the structure and principle of the microwave tomography system.The microwave transmission path is different from the linear propagation mode of Xrays.It will be affected by the electromagnetic field during the propagation process.Therefore,the finite element software is used to model and simulate the microwave Transmission path.The MWT system sensor model composed of 16 antennas and the cross-section of the object to be measured are built through multiphysics coupling software.The finite element method is used to solve the MWT system forward problem,and the scattered field data and sensitivity matrix for reconstruction are obtained.The matrix has been processed with mean value filtering to improve the uneven distribution of the sensitivity matrix.The scattering field projection data,sensitivity matrix and image gray value are normalized to improve the calculation efficiency of the reconstruction algorithm.Secondly,aiming at the singularity and ill-condition of the Jacobian matrix of the Newton iterative algorithm,and the distortion of the convergence direction,a reconstruction method combining the regularized Newton-conjugate gradient algorithm is studied.Quantitative and qualitative analysis of the reconstructed cross-sectional image and the object model to be measured using the image quality evaluation method verifies the effectiveness of the reconstruction algorithm.For each step of the Newton iteration method,the Jacobian matrix needs to be calculated,and an improved inaccurate Newton algorithm for cross-sectional reconstruction is studied.Only approximate solutions are needed.This can effectively solve large-scale nonlinear systems.At the same time,it can improve by monitoring the constant control sequence.Its local convergence and finding the global optimal solution make the inverse problem of microwave tomography more stable.Finally,combined with the actual environment,by adding Gaussian noise,the MWT image reconstruction algorithm is studied,and the cross-sectional model of the object to be measured and the reconstructed image are qualitatively and quantitatively analyzed by simulation experiments.Experiments show that the accuracy of the reconstruction algorithm studied in this paper has been better improved,and the iteration time consumption is not significantly improved,but under the premise of ensuring the speed,an effective balance of accuracy and speed has been achieved.