Street view imagery-based method for reconstructing 3D building façade openings

Abstract

The availability of 3D building models has been increasing, but they often lack detail at the architectural scale. This paper presents a method for reconstructing façade openings in 3D building models by integrating Street View imagery (SVI). Methodologically, the paper advances opening reconstruction in two key ways: first, by introducing a mathematically derived method for estimating unknown intrinsic camera parameters, enabling metric 2D-to-3D projection without relying on multi-view imagery or pre-existing depth information. Second, the method extends single-image photogrammetry to accurately measure detailed façade openings, converting pixel coordinates into spatial coordinates. The proposed method is validated through case studies in Amsterdam. Quantitative evaluation using the Façade Re-projection Dice Score (FRDS) shows high spatial consistency between reconstructed openings and reference opening geometries, with most scores ranging from 0.84 to 0.98. Given the broad coverage of SVI, there is a significant potential for enhancing 3D city models in diverse urban contexts where current representations remain geometrically basic.

Publication
Automation in Construction
Rui Ma
Rui Ma
Visiting Scholar
Filip Biljecki
Filip Biljecki
Assistant Professor