Hybrid Reconstruction Of Subsurface 3-D Objects Using FRTM And VBIM Enhanced By Monte Carlo Method

Ground penetrating radar(GPR)is widely used in the detection experiments of near-surface objects,and has the advantages of non-destructive,high resolution,and fast detection speed.It is also commonly used in highway crack detection,building defect detection,and geological exploration in actual life.For complex objects and their relative permittivity,the full waveform inversion method can be used to calculate and obtain accurate results.In this paper,starting from the frequency domain reverse time migration algorithm,the use of full waveform inversion combined with Monta Carlo method to carry out related research.Reverse-time migration imaging can clearly reconstruct the geological structure of the near surface and the geometric shape of the target object,and is suitable for judging the positioning and approximate shape of small objects in a large space.In practical applications,the computational efficiency and computational cost of reverse-time migration imaging are important criteria.In this paper,frequency domain reverse time migration imaging is used in imaging work.The principle of frequency domain reverse time migration imaging is briefly described.The same or even more accurate imaging results with other imaging algorithms are obtained with the same accuracy.This method determines The approximate position and size of the target object.Based on the above imaging results,full-wave inversion is used to simulate the specific shape and dielectric constant of the object,and the variational Born iterative method is used to iterate during the inversion process.The variational Born iterative method has the advantage of faster convergence rate than the Born iterative method,and has the advantage of less calculation amount compared to the modified Born iterative method.Thus the correct shape of the object and its electromagnetic parameters are obtained.Next,the statistical probability method Monte Carlo method is introduced,which determines the probability of occurrence of events in large-capacity random samples.Applying this method to compress the variational Born iterative space.Under a reasonable threshold,a part of the background space is eliminated by performing probability statistical judgment after each inversion iteration,and the inversion space is gradually compressed to a certain extent.The inversion rate promotes the variational Born iterative method to obtain more accurate inversion results.Finally,we simulate the joint imaging and inversion in the actual environment,add three different sizes of noise to the simulated data samples,perform imaging and inversion separately,and calculate the residuals of the final inversion data results and test data results.It is proved that the error can be reduced to a sufficiently small value in the inversion iteration.At the same time,the background dielectric constant is also changed to have a certain error with the background set by the model.The final result proves that the inversion result residual is highly sensitive to the error of the background dielectric constant.