Contribution of ultra-wide band and polarization diversity for the non-destructive evaluation of civil engineering structures using the ground penetrating radar (GPR)

The Ground Penetrating Radar technique (GPR) is now widely used as a non destructive probing and imaging tool in several civil engineering applications mainly concerning inspection of construction materials and structures, mapping of underground utilities and voids, characterization of sub-structures, foundations and soil and estimation of sub-surface volumetric moisture content. GPR belongs to a continuously evolving field due to electronic integration, high-performance computing, and advanced signal processing. The promotion of this technology relies on the development of new system configurations and data processing tools for the interpretation of sub-surface images. In this context, the work presents first the dual polarization UWB ground coupled GPR system which has been developed recently. Then, the data processing has focalized on the development of analysis tools to transform the raw images in a more user-readable image in order to improve the GPR data interpretation especially within the scope of detection of urban pipes and soil characterization. The processing means used concern clutter removal in the pre-processing step using adaptations and extensions of the PCA and ICA algorithms. Moreover, a template matching image processing technique is presented to help the detection of hyperbola within GPR raw B-scan images. The dual polarization is finally shown to bring additional information and to improve the detection of buried dielectric objects or medium discontinuities. The performances of our analysis approaches are illustrated using synthetic data (3D FDTD simulations) and field-measurement data in controlled environments. Different polarization configurations and dielectric characteristics of objects have been considered. The potential for target discrimination has been quantified using statistical criteria such as ROC