Automatic construction of geometric digital twins of existing buildings from point cloud datasets

Abstract

Digitising the geometry of existing buildings remains expensive due to the high amount of manual work necessary to process raw data. This indicates the demand for automatic solutions for geometry digitisation. This paper presents a few methods to detect and model structural objects and relations between objects and surfaces in large-scale occluded point clouds. We show that these methods give promising results and manage to capture the majority of the target entities.