Research On The Algorithm Of Electromagnetic Imaging And Inversion In Layered Media

Electromagnetic imaging and inversion algorithms are important research directions for electromagnetic detection.They can be used to reconstruct physical information of underground structures and detect the physical properties of media.They have important application value in practical engineering.For a long time,electromagnetic imaging and inversion algorithms are difficult to be applied in practical engineering because of the problems such as long computational time and large memory consumption.Therefore,it is important to study simplified and fast imaging and inversion algorithms.In view of the above,the research content of this paper contains two parts:1.A combined algorithm of background dielectric constant inversion and frequency domain reverse time migration imaging;2.A three-dimensional electromagnetic inversion algorithm based on deep learning.Reverse time migration imaging has been widely studied because of its advantages such as high imaging resolution and simple imaging conditions.However,reverse-time migration imaging algorithm in time-domain needs to know the electromagnetic parameters of the inversion area in advance,so as to determine the velocity model of electromagnetic wave propagation.The time-domain finite element method is used to calculate the wave field distribution in the inversion area,which takes a long time.Therefore,this paper will first propose a joint algorithm combining background relative permittivity inversion and frequency domain reverse time migration imaging.Relative permittivity of the background is first retrieved and then imaged.In order to verify the reliability and effectiveness of the inversion and imaging algorithms,this paper will use simulation data and experimental data to verify the algorithm.The traditional three-dimensional electromagnetic inversion first uses the receiver to receive data,and then uses the received data to solve the cost function to obtain the electromagnetic parameter distribution in the inversion area.This process is a typical inverse scattering problem and can usually be converted to a nonlinear or linear problem to solve.However,the cost of the solution process is generally very large,and the results of the solution may also have multiple solutions,resulting in a large difference between the final inversion results and the actual model.Therefore,this paper will propose a threedimensional electromagnetic inversion algorithm based on deep learning.The algorithm first obtains a preliminary image by Bonn approximate imaging,then uses the Monte Carlo method to remove the background noise,and finally uses the convolutional neural network to repair and accurately restore the electromagnetic distribution.In order to verify the reliability and superiority of our algorithm,we will design different test models and use the algorithm to obtain the inversion results.At the same time,the inversion results are compared with the inversion results of the variational Bonn iteration method(VBIM).