AI-assisted semantic segmentation for automated asphalt crack detection

The inspection of asphalt sealing hydro structures is time-intensive, safety-critical, and partly subjective, particularly in areas which are difficult to access. This article presents an AI-assisted workflow for automated crack detection and semantic segmentation on asphalt sealing surfaces at the Goldisthal and Markersbach pumped-storage plants in Germany. It was found that AI-assisted segmentation can support safer, more repeatable inspections, while providing georeferenced outputs suitable for traceable condition monitoring of hydropower infrastructure.