Scan2BrIM: IFC Model Generation of Concrete Bridges from Point Clouds

Generating a semantically enriched solid model of a structure from point cloud data is time-consuming. Many reinforced concrete bridges consist of a specific set of structures for which a large part of the processing from scan to bridge information model ( BrIM) can be automated with today's algorithms. We contribute a pipeline for going from a laser scan point cloud to a BrIM for simple concrete bridges ( girder, box girder, and slab). The procedure consists of four major steps: ( 1) top-down partitioning of the bridges, ( 2) bottom-up segmentation of the component surface elements, ( 3) recognition of components from surface elements, and ( 4) reconstruction of the components. The top-down partitioning step informs machine learning methods for identifying the bridge component type and its geometric model. Application of the pipeline to two bridges demonstrates that conversion to BrIM at level of detail 200 is possible. The pipeline takes less than 1 hour of user time, which is less than scanning time, thus inverting the Scan2BrIM labor ratio.