Automated 3D reconstruction of high-resolution images for tunnel inspection and monitoring purposes

The paper presents an automated tool for reconstructing a three-dimensional (3D) model of tunnel infrastructure using high-resolution images. In the last decades, image acquisition and survey with mobile systems, i.e. , sensors mounted on vehicles, have seen major advancements in the field of infrastructure inspection. Traditionally, image mosaicking has been addressed by means of photogrammetry approaches, but this methodology has not been suitable for specific high-resolution image data acquired with an innovative mobile mapping system. The developed methodology assigns spatial coordinates to each pixel of the acquired images, allowing the creation of a textured 3D geometric model of the tunnel surface. To this aim, the cross-sectional geometry of the tunnel was first parametrised using best fitting methods, simulating the tunnel shape as an ellipse in a proper coordinate system. The resulting 3D reconstruction from imagery offers a comprehensive and highly detailed representation of structural elements, enabling direct identification and evaluation of concrete defects, visible alterations, and millimetric cracks. This procedure facilitated the assessment of crack dimensions and their potential impact. The study opens new opportunities for the digitalization of tunnel inspection practices. The 3D scaled high-resolution imagery can enable easier detection, mapping and quantitative classification of defects. The obtained digital model, combined with other techniques, e.g. , AI detection algorithms and real-time monitoring data, can provide a comprehensive assessment of the actual state of tunnels. Finally, the method supports temporal monitoring, allowing the comparison of subsequent surveys to observe structural changes over time.