Ground Penetrating Radar Signal Parameters Inversion Based On Sparsity Constraint

Has been developing rapidly in recent years,the national transportation infrastructure,roads,railways and the scale of the subway tunnel increased significantly,to meet the demand of the quality detection and maintenance,it is urgently necessary to develop a ground penetrating radar(GPR)data automatic interpretation framework especially in the aspect of underground medium parameters inversion,Technicians adjusted parameters artificially through simplified formula or work experience to analyze the GPR signals in the past,which led to not exact inversion results.Simultaneously,the traditional methods are frequently applied in the monolayer structure.Hence,it is meaningful to research how to establish a systematic procedure for analyzing the GPR echo signal in multilayered structures.Based on this,the electromagnetic wave propagation model in multilayered structure,the delay and reflectivity estimation of the radar echo signal,and the parameter inversion are discussed in this paper.According to electromagnetic wave propagation rule,an electromagnetic wave propagation model with full consideration of noise and multiple reflections is constructed.With generalized reflection coefficient sparse,sparse decomposition techniques are used in GPR signal processing,signal delay and the reflection coefficient are estimated based on the sparse decomposition technology such as OMP(Orthogonal Matching Pursuit),StOMP(Sparsity Adaptive Matching Pursuit),SAMP(Sparsity Adaptive Matching Pursuit),VssAMP(Variable step size Adaptive Matching Pursuit),and the linear inverse convolution and deconvolution method L2 estimation results were compared.Experimental results show that when echo signals do not overlap,both the deconvolution and sparse decomposition methods could estimate the radar signal delay and reflectivity under the environment with the jamming of echo and noise,but the deconvolution consumes more time;when echo signals overlap,the sparse decomposition technique is still able to estimate the delay and reflectivity accurately,while the deconvolution method could estimate the delay precisely but not estimate reflectivity exactly because of the influence of multiple echoes,which would make a bigger error in the later parameter inversion results.On the cornerstone of realizing the signal delay and reflectivity estimation,the medium parameters are inversed in layers by combining the previously established electromagnetic wave propagation model.Parameter inversion results demonstrate that the inversion precise of OMP is higher than ones of other sparse decompositions.However,the OMP algorithm is executed relying on the signal sparse degree,while the parameters in multilayered structures can be reflected accurately by StOMP,SAMP and VssAMP in the case of unknown the signal sparse degree.The original signal is approximated adaptively by changing the step size in VssAMP,which takes advantages of StOMP and SAMP.Therefore,the most accurate inversion results and the shortest time-consuming are obtained.And practical application could references this method.