Research Of The Process Of GPR Data On Stratified Media Based On Internal Reflections

In the field of civil engineering, geophysics Radar (Ground Penetrating Radar, GPR) is used as a tool for nondestructive testing (NDT) to probe the materials and structures of civil engineering (structures, pavement and soil). The radar antenna sends waves toward the material, and if the dielectric contrast is big enough, each interface of the medium generates a reflected wave. The aim of signal analysis is then to estimate the time delays of the echoes, which infers the geometric structure of the subsurface.In this report, we consider for a simple layered medium, which sought to determine the structure. The conventional signal model only includes the direct echoes (or primaries), which encounter only one reflection at the interfaces of the medium. The echo amplitude can be deduced from the expression of the reflection and transmission coefficients with Fresnel relations.Compared with the latter conventional model, we consider in this report another signal model which takes account for multiple echoes. As opposed to primaries, multiples echoes encountered several internal reflections between interfaces. Compared to the data model in Liu Limei, this report incorporates some further modeling on the echo's amplitude, i.e., geometric progression. This model introduces some non-linearity in the frequency behavior of the signal model, then simulation results of TDE on the new signal model will be given.There are various TDE (Time Delay Estimation) techniques in the literature. The FFT (Fast Fourier Transform) is the conventional method and has been widely used due to its simplicity and high efficiency for separating echoes. But there is some limitation, FFT can not work when two signals are very close to each other. For applications requiring a better time resolution, we can use methods such as the high resolution algorithm MUSIC (MUltiple SIgnals Classification).MUSIC algorithm can not work when the echoes are correlated. This is not realistic in part of our application because the echoes originate from the same source. To improve the performance of MUSIC within such a multipath environment, a preprocessing is introduced, i.e., (MUSIC-SSP, MUSIC-MSSP). It is based on the data averaging over overlapping frequency subbands in order to reduce the cross-correlation magnitude between echoes.In this report, we first explain some applications that are based on the inverse Fourier transform with the signal model. Then, the operating principle of the MUSIC algorithm is introduced. And then, to enhance the ability of the MUSIC, we present the pre-processing techniques by averaging sub-bands. Finally, simulation results are given as well a conclusion and perspectives.