Research On Inversion Algorithm Of Layered Media Parameters Based On Ground Penetrating Radar

Ground penetrating radar(GPR)has the characteristics of non-damage,high efficiency,accuracy and portability,and can accurately obtain the structural characteristics of a variety of underground media,so it is widely used in the field of layered media detection.In order to accurately depict the structure of layered media,it is necessary to accurately inverse the parameters of each layer.However,the GPR echo data is often accompanied by noise interference,which will misjudge the target position,resulting in inaccurate inversion of medium parameters.The existing horizon tracking technology will appear the phenomenon of "string layer" at the interface of layered media with great fluctuation,which will lead to the inaccurate calculation of two-way travel time and reduce the inversion accuracy of media parameters.In addition,the traditional inversion algorithm optimizes the objective function within the randomly generated initial range,and the initial value far from the real value is repeatedly calculated,which will reduce the efficiency of inversion.The number of inversion parameters increases with the increase of the number of media layers,the inversion algorithm is easy to fall into the local optimal value,which will lead to low inversion accuracy.In view of the above problems,in order to improve the efficiency and accuracy of layered media parameter inversion,this paper improves the inversion algorithm from three aspects:suppressing echo data noise,accurately calculating two-way travel time and optimizing inversion algorithm.In order to solve the problem that the noise in the echo data will cause the misjudgment of the target position,and then lead to the inaccurate inversion of medium parameters,a GPR data noise suppression method based on dictionary learning is proposed.This method makes use of the sparse representation of the target reflection echo of the layered medium in the dictionary to effectively distinguish the target signal from noise.However,when the noise environment changes,the orthogonal matching pursuit(OMP)algorithm used in the sparse coding phase of the dictionary learning method will mistakenly choose the basis function,resulting in incomplete noise removal in the echo data,thus reducing the accuracy of inversion.Therefore,the fast iterative shrinkage threshold algorithm(FISTA)is used to sparsely encode the echo data,so that the iterative threshold can be adjusted according to the characteristics of the echo data.At the same time,the dictionary is updated with K-SVD algorithm to meet the denoising requirements under different noise conditions.The simulation results show that the algorithm can improve the signal-to-noise ratio of echo data and restrain the influence of noise on the accuracy of medium parameter inversion.In order to solve the problem that the horizon information obtained by the existing horizon tracking technology is not accurate,and then reduce the inversion accuracy of media parameters,a two-way travel time calculation method based on horizon tracking is proposed.Because the width of the sliding window of the traditional horizon tracking algorithm is fixed and there is no error correction mechanism,the edge detection algorithm of Canny operator is used to pick up the horizon range and determine the boundary of the sliding window.Secondly,the horizon tracking technology with three-level window is used to prevent the real seed from falling on the outside of the window.The simulation results show that the horizon information extracted by the improved horizon tracking algorithm at the undulating interface is closer to the real value,and the calculated two-way travel time information is more accurate.In order to solve the problem that the inversion algorithm is easy to fall into local optimization and has low computational efficiency,an inversion algorithm based on adaptive weight-resampling-particle swarm optimization(A-R-PSO)is proposed.Firstly,the initial values of the parameters of layered media are estimated by using the amplitude spectrum and phase spectrum of the generalized reflection coefficient respectively,which reduces the range of initial values of particles.Secondly,the resampling algorithm is used in the process of particle iteration,which eliminates the particles far from the real value and saves computing resources.Finally,the adaptive inertia factor is introduced into the speed iteration formula of the PSO algorithm to balance the local search ability and the global search ability,so that the particles converge to the global optimal solution faster.The simulation results show that the improved inversion algorithm improves the efficiency and accuracy of inversion.