Research And Implementation For GPR Forward Modeling And Migration Algorithm

In the practical engineering, the Ground Penetrating Radar data is interpreted mainly relying on the experience of GPR expert or manual analysis methods on image, which is highly subjective and often leads to inaccurate interpretation of GPR data, then affecting the operations of geological exploration. In addition, the training of GPR expert is also a time-consuming and labor-intensive process. GPR forward modeling technique can not only improve the accuracy of GPR data interpretation, but also shorten the training period of GPR expert. However, due to the influence of special recording mode of GPR and wave propagation characteristics, the result of GPR forward modeling is only a rough estimation of subsurface medium, which can be improved by migration algorithm. The resolution of GPR forward image can be improved by migration processing, so that make GPR image more realistically reflect the geological structure. However, the processing effect of current migration algorithm is only subjective improvement of vision and lack of objective quantitative criterion of evaluation; the choice of migrating velocity depends largely on personally experiential knowledge; besides, the filtering of direct wave is also one of the key factors to achieve the objective and accurate evaluation of GPR image.Aiming at these problems, a series of algorithms are researched in this paper, which achieve the GPR forward modeling, migration processing and GPR image evaluation more systematically, mainly including as following:(I) FDTD forward modeling:it starts from Maxwell's curl equations in the frequency domain, according to the FDTD principle with the absorbing boundary condition based on convolutional perfectly matched layer(CPML), and the forward modeling simulation on two-dimensional geological model is performed, by which the receiving data of GPR simulation is gotten, namely the forward data profile.(II) Research on migration algorithm:the migration algorithm is used to disposed the GPR forward profile, which can make the reflecting wave homing and diffracting wave energy converging in forward profile through comparing the results before migration with the ones after, thereby enhancing the resolution of the forward profile. Simultaneously, the processing performance of three used kinds of migration algorithms based on wave equation is comparatively analyzed.(Ⅲ) Image evaluation and filtering in GPR migration processing:aiming at the problem existing in migration processing, the image evaluation method based on minimum entropy technique is educed. Combining with it, the general velocity model suitable for point-like target migration and the segmentation velocity model suitable for extended target migration are presented respectivel, by which the objective and quantitative evaluation of migration processed image is achieved and the migration imaging of subsurface abnormal target and the inversion of background permittivity and target size as well. Finally, through comparing the filtering performance of F-K filter with that of mean filter, the feasibility of F-K filter to filter out the direct wave is discussed.