Research On Inverse Scattered Microwave Imaging Of High Contrast Medium Targets

The electromagnetic inverse scattering problem is the problem of retrieving the spatial position and material properties of the target based on the measured electromagnetic scattering field data.It has a wide range of applications in biological imaging,nondestructive evaluation,wireless communication,and security screening.The main challenge in solving the electromagnetic inverse scattering problem is its inherent ill-posed and nonlinearity.Since the number of unknowns to be solved is much larger than the number of equations,the inverse scattering problem is an ill-posed problem.On the other hand,the nonlinearity of the unknown constitutive parameters is due to the multiple scattering effect inside the region.Generally speaking,the higher the contrast of the target medium,the stronger the multiple scattering effect it brings.When dealing with highcontrast medium targets,the multiple scattering effect is significant,corresponding to a highly nonlinear inverse scattering problem.Aiming at the problems of low inversion efficiency and poor imaging accuracy of traditional methods in the inversion process of high-contrast medium targets,this thesis studies an effective solution to the highly nonlinear inverse scattering problem.The main research content of this thesis consists of two parts.The first part starts from the deterministic inversion theory,and firstly studies the systematic structure of the inversion model.A new inversion model is constructed by reasonably changing the variable frame,deductive form,weight coefficient and other factors in the objective function through physical cognition or empirical laws.The second part of this thesis proposes a fast firebug swarm optimization algorithm based on deep learning.By combining the stochastic optimization algorithm with the deep learning multi-resolution strategy,the computational burden of the stochastic optimization algorithm to solve the inverse scattering problem is effectively relieved.The main innovations of this thesis are summarized as follows:1.Deterministic method for low nonlinear structures based on generalized contracted integral equation model.Based on the contracted integral equation model for highly nonlinear inverse scattering problem,a generalized contracted integral equation model is proposed from the perspective of deterministic optimization theory.Compared with the original contracted integral equation model,it can further reduce the nonlinearity of the model structure of the inverse scattering problem.On this basis,a new inversion imaging method is proposed,which consists of the following three parts:the algorithm framework of contrast current source based on fast Fourier transform,generalized contracted integral equation model and weight adjustment nested iteration scheme.2.A fast algorithm for firebug swarm optimization using deep learning.Aiming at the highly nonlinear inverse scattering problem,starting from the perspective of stochastic optimization theory,a fast algorithm for firebug swarm optimization based on deep learning is proposed,and the specific algorithm flow of its realization is elaborated.The innovation of this algorithm is to reduce the computational burden of the stochastic optimization algorithm to solve the inverse scattering problem by exploring the use of deep learning capabilities and image interpolation technology.This is different from the existing learning assistance methods based on various gradient-dependent solvers,because compared with the deterministic optimization algorithms that strongly rely on the initial solution,the stochastic method can jump out of the local optimum and has a better global optimization ability.