Researchers at Seoul National University of Science and Technology have developed an artificial intelligence framework that can monitor bridge damage over time using drone images collected during routine inspections.
The system combines computer vision, drone imagery, three-dimensional bridge reconstruction, and AI-based image alignment to track how cracks, concrete spalling, and water leakage change across multiple inspections.
Bridges are critical parts of transportation networks, but they gradually develop damage due to traffic loads, weather conditions, aging, and environmental exposure. Regular inspection is essential for public safety, but traditional visual inspections can be time-consuming, expensive, and risky for workers.
Computer vision has made automated inspection more practical, but long-term monitoring remains difficult. Images taken months apart are often captured from different angles, distances, and positions, making it hard to compare the same damage accurately over time.
To address this challenge, a research team led by Assistant Professor Hyunjun Kim developed a framework that uses drone images to build a three-dimensional reference model of a bridge during the first inspection.
For later inspections, new images are automatically aligned with the original 3D bridge model using hierarchical localization and image clustering. This allows the system to recognize and compare the same damaged areas even when the photographs are taken from different viewpoints.
The framework also uses Global Navigation Satellite System data to convert image-based measurements into real-world dimensions. This makes it possible to estimate the actual size of damaged areas and track their progression over time.
The researchers validated the system over 120 days using drone imagery from an in-service prestressed concrete bridge. During the monitoring period, the AI framework successfully tracked cracks, spalling, and water leakage despite changes in camera viewpoint.
The system measured damaged areas with a maximum error of only 4.61% compared with conventional manual measurements. This shows that the method can deliver accurate long-term damage tracking while reducing reliance on repeated manual inspection.
Unlike traditional methods that mainly assess damage at a single point in time, or require new 3D models for every inspection, the new approach uses one reference model for continuous monitoring. This improves consistency and reduces the computational effort needed for long-term bridge assessment.
Although the current method works best on relatively flat structural components and may need further improvement for highly curved surfaces, it is suitable for many bridge elements commonly inspected in routine maintenance.
The researchers believe the framework could help engineers make better maintenance decisions, extend the service life of bridges, and reduce long-term inspection and repair costs.
As transport agencies face growing challenges from aging infrastructure, AI-powered monitoring could support predictive maintenance strategies and improve public safety. In the future, the same approach may also be adapted for tunnels, dams, elevated rail systems, and other critical infrastructure.
Journal Reference:
Kim, H. (2026). AI-powered long-term bridge damage monitoring using drone inspections. Structural Health Monitoring. https://doi.org/10.1177/14759217261443618