Introducing a Novel Protocol for Collecting Annotated Images to Automate Concrete Structure Damage Severity Assessment

Concrete structures like dams, bridges, and buildings require regular inspections to ensure public safety. Currently, these inspections are carried out manually, which demands significant human and financial resources and poses risks to inspectors due to accessibility challenges. A more promising approach involves automating inspections using drones and algorithms to detect and estimate cracks. This paper introduces a protocol for capturing high-quality images of concrete structures using a tripod and a high-resolution camera. The collected images were used to develop a robust algorithm to estimate crack widths and classify them based on severity according to the Manuel d'inspection des structures du Ministère des Transports et de la Mobilité Durable du Gouvernement du Québec. The algorithm's accuracy was validated by comparing its measurements with actual crack values obtained directly from the inspected structures. Additionally, the new database, now accessible to the public, contains numerous measurements that could prove invaluable for future research endeavors, inspiring breakthroughs in structural health monitoring.