Automated bridge damage mapping from UAV imagery slicing with BIM-based enrichment

This study proposes an automated pipeline for bridge damage detection and mapping using UAV imagery integrated with Bridge Information Modelling (BrIM). The framework combines a YOLOv11x-seg architecture with Slicing Aided Hyper Inference (SAHI) to preserve fine details in high-resolution images, and employs a hybrid segmentation strategy with the Segment Anything Model (SAM) for class-specific refinement, achieving an overall mAP50 of 0.593. Identified damages are projected onto 3D meshes through multi-view ray casting, where adaptive thresholding and clustering isolate individual defects. Orthophoto-based quantification then provides accurate measurements of crack geometry, exposed rebar, and efflorescence areas, strictly bounded by a spatial resolution limit of 1.0 mm/pixel. The results are embedded into Industry Foundation Classes (IFC) models and linked to BIM Collaboration Format (BCF) reports, ensuring interoperability and supporting lifecycle management. Demonstrated on three Portuguese railway bridges, the pipeline delivers a scalable and verifiable “as-damaged” BrIM, advancing predictive maintenance and structural asset management.