| The step frequency system ground penetrating radar(GPR)is an emerging ground penetrating radar system in recent years,and its biggest advantage is that it can consider the requirements of detection depth and target resolution.It has a larger bandwidth and is more suitable for detecting shallow subsurface targets.The sparse Bayesian learning(SBL)method is an important branch of sparse representation theory.It uses prior information of the sparse signal,combined with Bayes’ theory to obtain posterior information of the signal in order to solve the unknown parameters.The introduction of SBL into the field of GPR imaging can not only reduce data storage and data acquisition costs but also significantly improve noise immunity.Accordingly,this paper combines SBL with the step-frequency continuous-wave GPR to further investigate the GPR imaging problem.The work in this paper is summarized as follows:1.For the traditional sparse Bayesian algorithm does not consider the structural characteristics of ground penetrating radar signals leading to the subsurface target imaging accuracy is not high and only suitable for real number field problems,a complex block sparse Bayesian compressed sensing imaging algorithm is proposed.In this algorithm,the reflection coefficient of the target is reconstructed by establishing the sparse Bayesian model and applying the complex Gaussian scale hybrid model,which extends the block sparse Bayesian learning model from the real field to the complex field.Block sparse model can utilize the structural characteristics of GPR signals to block sparse signals and dictionaries to improve reconstruction accuracy.The GPRMAX simulation software is used to build the scene of GPR,obtain the time-domain data,and reconstruct the position information of the subsurface targets.The experimental results show that the accuracy of the proposed algorithm is improved compared with other sparse reconstruction algorithms,and the imaging effect is better at low signal-to-noise ratio.2.To address the problems of large data volume and long computation time required for traditional ground-penetrating radar 3D imaging,a method for imaging GPR with dimensional sparse Bayesian is proposed.The method first makes full use of the relationship between MIMO array elements,decomposes the 3D dictionary matrix into three sub-matrix forms,where each sub-matrix contains one dimensional information of spatial points.Then,sparse Bayesian imaging method is performed on each sub-matrix of each dimension to obtain a complete 3D image.As the dictionary is decomposed,the computational volume is greatly reduced and the computation time is shortened.It is experimentally demonstrated that the proposed algorithm achieves fast imaging with reduced imaging time compared with the standard sparse Bayesian imaging method and OMP fast imaging method3.For the problem of inaccurate GPR imaging positions in multilayer media,a sparse Bayesian 3D GPR imaging algorithm that can be applied in multilayer media is proposed.In the case of multilayer media,the selection of refraction points in the traditional method will have a large impact on the accuracy of subsurface target imaging.Therefore,an equivalent media layer method is adopted to simplify the multilayer media model,and the sparse Bayesian algorithm is used to reconstruct the model is reconstructed.The experimental results show that the imaging position of the subsurface target is more accurate and the imaging results are more intuitive under this method. |
