AI-powered bridge defect detection and condition assessment using drone-collected visual data

Abstract Maintaining safe bridges requires regular inspections, but traditional methods depend on manual visual surveys that are labor-intensive, subjective, and often limited by access and safety constraints. Recent advances in drones and artificial intelligence (AI) provide opportunities to improve efficiency, consistency, and coverage in bridge monitoring. In this study, we demonstrate a workflow that uses drone collected data to detect and segment eight types of concrete deterioration. Post-processing methods stabilize detections across video frames and produce machine-readable outputs that align closely with manual tallies. Results from ten bridges in Kansas indicate that the automated defect counts show strong directional consistency with the nationally reported condition ratings. The approach also projects detections into three-dimensional reconstructions for measurement and visualization, and a prototype web service enables near real-time deployment. Collectively, these findings suggest that drone- and AI-powered inspections can effectively complement traditional inspection practices while offering scalable and data-driven support for infrastructure management.