LiDAR health monitoring of transport infrastructure: automatic damage detection using classification of point clouds

Health monitoring of transport infrastructure plays a central role in ensuring the efficient management, the safety and functionality of transport systems with impacts on the economy and the development of entire countries. In this context, particular attention is dedicated to bridges, following recent dramatic collapses occurred worldwide. This is also the case of Italy, where in accordance with current bridge inspection guidelines at the European level, specific guidelines for the classification and risk management, safety assessment, and monitoring of existing bridges were issued, with a particular focus on regular inspections for identifying potential damages and structural issues. Traditional monitoring procedures are based on visual on-site inspections conducted by specialized operators, which have the advantage of acquiring data with high accuracy, useful for planning both routine and extraordinary maintenance interventions. However, these methods are characterized by low repeatability on time, associated to high costs, and could be also dependent to the experience of the operator. To overcome these limitations, recent advancements in Non-Destructive Testing methods (NDTs), particularly those utilizing LiDAR (Laser Imaging Detection and Ranging) technologies, have opened new avenues for transport infrastructure and bridge assessment. This study explores the application of point clouds, collected by Terrestrial Laser Scanner (TLS), for automated defect identification in transport infrastructure, with a specific focus on signal amplitude to enhance the process. The proposed approach allows the automatic identification of potential damages, based on the classification and segmentation of cloud points collected by a TLS. To this purpose, an algorithm was applied for clustering the point clouds into uniform regions, enabling an accurate analysis of structural potential damages. The primary innovation involves the integration of signal amplitude. It reflects material variations and surface defects with a millimeter accuracy, into the segmentation process. This integration allows for more precise identification of structural anomalies, such as cracks, discontinuities, or corrosion, through an analysis of amplitude variations. Experimental results were conducted on a case-study demonstrating the effectiveness of the proposed approach in automatically detecting damaged areas or defects, leading to a significant reduction in on-site inspection time. Furthermore, the automatic defects identification process streamlines the entire inspection workflow and opens the way for a more efficient method providing a comprehensive defects map within the georeferenced points cloud.