Research On Efficient Reconstruction Algorithm Of 3-D Objects In Layered Media

Electromagnetic inverse scattering problem is a kind of mathematical inverse problem or known as electromagnetic inversion.It uses the scattering field measured outside the object to infer the position,shape,dielectric properties and other information of the object.Related electromagnetic inversion methods are widely used in geophysical exploration,biomedical imaging,through wall radar and other issues.For a long time,3D electromagnetic scattering and inverse scattering are difficult to be applied in practical engineering because of the problems of long calculation time and large memory consumption.On the other hand,with the wide application of anisotropic materials in microwave equipment,optical components and aviation industry,the inversion of 3-D arbitrary anisotropic objects puts forward higher requirements for the existing algorithms,so the fast and stable 3-D inversion algorithm is the key research direction in the industry.1.The combination of unsupervised learning and 3D isotropic irregular object inversion method;2.A cascade solver is used to reconstruct arbitrary anisotropic objects in layered media.In the first part,we use voxel based variational Born iterative method(VBIM)combined with an unsupervised learning algorithm---Maximum Entropy Threshold Segmentation Technology(MET)to reconstruct 3-D irregular objects embedded in layered media.In each iteration,the model parameters in all discrete meshes in the inversion region are reconstructed by using the advantages of voxel based inversion,and then each output mesh is classified as "background" and "scatterer" by MET.By removing the "background" mesh which has no contribution to the measured scattering field,the dimension of unknown vectors in the discrete data equation is significantly reduced.Finally,through numerical experiments,compared with the traditional VBIM algorithm,it has significant advantages in computational efficiency,memory cost and anti-noise ability.The second part is about the inversion of three-dimensional arbitrary anisotropic objects,which has more extensive prospects and engineering needs in practical applications.However,the inversion of arbitrary anisotropic objects involves very high unknown vector dimensions,which often requires huge computational resources.In this part,we cascade three inversion methods,including isotropic inversion,biaxial anisotropy inversion and arbitrary anisotropic inversion,which are used to reconstruct multiple dielectric parameters of three-dimensional arbitrary anisotropic objects in layered media.The dielectric parameters obtained by the first stage solver will be used as the initial values of the next stage solver.At the same time,according to the differences between the inversion dielectric parameters and the background parameters,we synchronously narrow the inversion region.The numerical simulation results show that the cascade inversion solver can not only obtain more accurate reconstruction results,but also significantly reduce the computational cost compared with the direct anisotropic inversion.In addition,we also test the anti noise ability of this method.Finally,we summarize the paper and think about the future work.