Automatic completion of geometric models from point clouds for analyzing historic timber roof structures

Laser scanners are being increasingly utilized in the engineering community to collect accurate and dense 3D data on timber roof structures. Point clouds, which are produced by either scanning or photogrammetry, can be subsequently processed using specialized software to create geometric models of individual components of the roof structure manually. Manual modeling of each structural element in a computer-aided-design-based structural analysis software application is a labor-intensive and time-consuming process, whereas automatic modeling typically yields incomplete results in terms of the numbers and sizes of individual beams owing to shortcomings in data acquisition or data processing. In the context of structural assessment, full-scale modeling of each beam is mandatory for a holistic analysis of structural health. This study addresses the gap between automated roof-structure models and preliminary structural assessments through a hierarchical analysis and refinement of the geometric model. The proposed method was applied to two different roof structures, and structural assessments show that it is feasible to import and process the generated models in a structural analysis software application. The resulting structural models indicate that the workflow improves the months of time consumed by optional mistakes from manual contributions to less than a week without the uncertainties caused by human error.