A flexible methodology for outdoor/indoor building reconstruction from occluded point clouds

Abstract

Terrestrial Laser Scanning data are increasingly used in building survey not only in cultural heritage domain but also for as-built modelling of large and medium size civil structures. However, raw point clouds derived from laser scanning generally not directly ready for the generation of such models. A time-consuming manual modelling phase has to be taken into account. In addition the large presence of occlusion and clutter may turn out in low-quality building models when state-of-the-art automatic modelling procedures are applied. This paper presents an automated procedure to convert raw point clouds into semantically-enriched building models. The developed method mainly focuses on a geometrical complexity typical of modern buildings with clear prevalence of planar features A characteristic of this methodology is the possibility to work with outdoor and indoor building environments. In order to operate under severe occlusions and clutter a couple of completion algorithms were designed to generate a plausible and reliable model. Finally, some examples of the developed modelling procedure are presented and discussed.