Ground Penetrating Radar Surveys To Surface Roughness Estimation

Ground penetrating radar (GPR) is a common tool for non-destructive probing in civilengineering. GPR plays a significant role in highway construction and maintenance. Forcentimeter-scale wavelengths GPR is mostly used for road pavement survey. For this propose,the roadway is assumed to be horizontally stratified. The vertical structure of the roadway canbe deduced form radar profiles by means of echo detection and estimation. Conventionalmethods used to estimate the structure of pavement assume flat interface, which bringdeviation in radar signal detection and processing, particularly in high frequency. Here theinfluence of interface roughness (surface and interlayer roughness) is accounted for within thescope of the data processing of radar signals. Joint estimation algorithms are used to estimateboth time delay and roughness parameters.The work shows as follows:1. A mathematical model is developed, which takes influence of interface roughness intoaccount. The rigorous electromagnetic method PILE (propagation inside layer expansion)provides simulated data that show the influence of the interface roughness on thebackscattered echoes of stratified media in the ultra-wide band (0.5GHz-10.5GHz). AndPILE is also used for studying the frequency behavior of reflected echo (like amplitude andphase), then curve fitting is made and a mathematical model of frequency behavior ofreflected echoes is developed.2. Joint estimation algorithms are presented. Layer thickness can be estimated by usingtime-delay estimation of different echoes calculated from PILE method. When layerthicknesses less than5cm, high-resolution algorithms or super high-resolution algorithmswill be used for improving the performance of GPR. After considering the influence ofinterlayer roughness, the observed frequency variations of the radar magnitude introducesome bias on time delay estimations, which is quantified with the subspace algorithm MUSIC.For solving this problem, MUSIC needs to be modified. De-correlation processing SSP andMSSP are used before new algorithms implement. Finally, an adaptation of MUSIC and2-DMUSIC allows us to jointly estimate the time delay and the interface roughness.In conclusion, PILE method can accurately calculate frequency behavior of reflected echoes from each layer, also Modified MUSIC algorithm could jointly estimate time delayand roughness parameter precisely. Parts of work were written into a paper and the paper wasaccepted by IWA GPR2013.