Subsurface imaging and characterization using full-wave inversion of near-field GPR data

Non-destructive characterization of the structural and physical properties of soils and materials has remained a major challenge in the scientific community. In that respect, ground-penetrating radar (GPR) has progressively become a key technology. The ever increasing computational performances available together with recent theoretical advances have made full-wave modeling and inversion a rational choice nowadays. Through full-wave inversion, information retrieval capabilities are inherently maximized. Approaching the radar antenna to the medium permits deeper characterization and higher resolution but also increases the complexity of the modeling problem. In this thesis, we further analyzed and advanced the processing of near-field GPR data. In particular, the impact of the number of frequencies and multi-offset antenna configurations on multilayered media reconstruction was investigated using objective function analyses. Focus was then dedicated to the development, optimization and validation of new radar data processing strategies. We introduced a novel numerical approach to remove antenna effects from the radar data, including near-field interactions with the medium, and, thereby, to provide much better radar images. This approach inherently normalizes the measurements, which makes straightforward the fusion of multi-frequency radar data for improved resolution. Finally, the near-field antenna effects removal and inverse modeling procedures were integrated in a GPR-based methodology to help detecting leaks in water distribution networks. This thesis opens new research and application avenues of GPR in civil and environmental engineering, e.g., for road inspection, among others.(AGRO - Sciences agronomiques et ingénierie biologique) -- UCL, 201