Drone-based photogrammetry and virtual reality: technological alternatives for tunnel inspection

Abstract

This study presents a comparative analysis of tunnel inspection methodologies, emphasizing the effectiveness of Virtual Reality (VR) inspection using drone-based 3D photogrammetric models in contrast to traditional visual inspection (TVI). The experiment was conducted in a 750-meter-long engineered concrete road tunnel in São Paulo, Brazil. The research involved fixing 139 synthetic pathology indicators inside the tunnel to standardize the assessments, and the photogrammetric model was compared to the Sacannig terrestrial laser – TLS 3D model, to evaluate the geometry quality. Results revealed that TVI achieved a precision of just 18% and an accuracy of 63%, highlighting significant subjectivity and inconsistency in manual inspections. In contrast, the VR method demonstrated an improved precision of 73% and an accuracy of 98%, with a reduction in the false discovery rate from 32% to 1%. The drone inspection was completed in approximately 35 min, significantly reducing inspection time compared to the two hours required for TVI. Furthermore, the photogrammetric approach proved better than Terrestrial Laser Scanning (TLS) in terms of occlusion reduction and visual realism, although TLS maintained advantages in data processing speed and accuracy. To the best of our knowledge, this is the first replicable study that directly compares VR-based inspection with TVI using a standardized dataset of synthetic tunnel pathologies, providing a robust ground-truth benchmark for future research. These findings underscore the potential of drone-based photogrammetry combined with Virtual Reality to enhance the efficiency, reliability, and scalability of tunnel inspections.