Automated extraction of elevation profiles from limited imagery for bridge design

Bridge design is a complex challenge influenced by topography, geotechnical conditions, and construction methods, with the elevation profile being crucial for determining structural typology and foundation design. This paper presents a method for automatically extracting bridge elevation profiles from a limited set of images, addressing data scarcity and semantic interpretation challenges. The workflow includes: (1) semantic segmentation of structural components, (2) image-based 3D scene reconstruction, (3) point cloud alignment and scaling, and (4) elevation profile extraction. Extracted profiles are evaluated against high-resolution 3D laser scans, achieving a mean cloud-to-cloud distance as low as 21 cm using only 12 images. Comparisons with public GIS data show superior accuracy for small bridges and, in an initial test, comparable results for larger structures. This approach enables efficient acquisition of topographical data from any available imagery, including public sources, reducing reliance on traditional surveying and potentially expediting future bridge design processes.