AI framework for bridge damage monitoring

Researchers at Seoul National University of Science and Technology, led by Assistant Professor Hyunjun Kim, developed an AI-powered computer vision framework that tracks and measures structural damage on bridges over time using routine drone inspections. The system anchors later inspections to a single 3D reference model built from initial drone images, using hierarchical localization and image clustering to align subsequent photos despite changes in camera angle or distance. It also incorporates Global Navigation Satellite System data to convert pixel measurements into real-world dimensions. The framework was validated over 120 days on an in-service prestressed concrete bridge, successfully tracking cracks, spalling, and water leakage with a maximum measurement error of 4.61% compared to manual methods. The approach reduces computational effort by reusing one reference model instead of rebuilding 3D models for each inspection, supporting predictive maintenance and extending infrastructure service life. The findings were published online in the journal Structural Health Monitoring on April 27, 2026. The method is best suited for relatively flat components, and the team suggests it could be adapted to tunnels, dams, and elevated rail systems.