Development of 3D visualisation algorithms for the effective interpretation of tunnel & transport infrastructure subsurface radar data

Abstract

Ground Penetrating Radar (GPR) is a versatile technology for non-destructive inspection of subsurface structural anomalies, which ensures the safety and e cacy of ageing transport infrastructure. E ective subsurface inspection better informs targeted maintenance strategies, maximising long-term infrastructure resilience. However, multiple research gaps exist for traditional xed-directional GPR. They extend to increasing the completeness of subsurface coverage (especially in fully-enclosed tunnel structures) and reducing reliance on unclear 2D visual output (based on hyperbolic signal artifacts); alongside addressing the lack of versatile, open-access data processing tools for 3D anomaly pro ling. This thesis advances emerging hybrid-rotational GPR technology, which utilise advanced 360-degree orientable antennas to maximise subsurface coverage. Work presented develops a unique 3D visualisation framework for this technology, focusing on the inspection of historic railway tunnels and retaining walls. The devised framework encompasses both data collection and conveyance strategies, returning practical 3D visual outputs for surveyors. Investigatory work systematically analyses framework performance across multiple infrastructure analogues. A novel, highly versatile 3D visualisation work ow is developed for e ective data conveyance, uniquely capable of pro ling subsurface anomalies in (i) simulated, (ii) xed-directional, and (iii) new hybrid-rotational GPR datasets. This work ow innovatively utilises exclusively open-access tools and a three-phase pipeline for data extraction, regularisation, and 3D segmentation. Innovations for data collection focus on developing next-generation hybrid-rotational GPR equipment for laboratory and eld testing. Key advancements include successful anomaly detection at decimetre scales to within 30mm, and consistent mid-centre localisation within 3% deviation from ground truth geometry. Combined Processing Methodology further increases imaging resolution, enhancing response pro le median frequency by 152%, sharpening amplitude peaks. Further high-impact research avenues are also identi ed, including automated anomaly characterisation, report generation, and dataset hybridisation. Overall, this thesis highlights the signi cant potential of the 3D visualisation framework for increasing transport infrastructure resilience, and develops essential resources to catalyse future innovation.