Arnaud SCHENKEL / Nadine WARZÉE
(Université Libre de Bruxelles, Brussels, Belgium)

Abstract: Rehabilitation or in-depth transformation of a site usually requires detailed accurate plans. Due to the state, the nature and the complexity of the site, it is sometimes difficult obtaining accurate plans with conventional techniques. The project purpose is the extraction of 2D plans from data acquired with 3D scanner devices.
A rehabilitation project aims to transform the bunker located in Brussels Park into a high-tech center for addressing new forms of warfare. As a vestige of World War II, it is a particularly appropriate location. Indeed, it directly evokes Belgian struggle in terms of protecting the population and the importance of telecommunication networks in wartime.
Because of the secrecy of the site, it is impossible to find complete plans that incorporate the changes over time. A preliminary topographic survey resulted in the production of plans, but with multiple inconsistencies in both geometry and topology.
A three-dimensional scanning campaign was conducted to satisfy two objectives: to obtain necessary measures to produce engineering drawings, and to allow conducting virtual visits of the building. The acquisition process in the field has resulted in a high-quality 3D model. The precision of the registration of eighty-one scans, despite the risk of loop closure problem, guarantees the quality of the whole model and therefore the possibility to extract correct plans.
For this, filtration and primitive fitting techniques have been applied to this massive data set, containing 180 million points. Our approach consists in a 2D transformation of the problem, to accelerate the process. A strip of the 3D model representing the walls is projected onto a flat surface. The results obtained can then be improved using techniques requiring less computing time and memory, as lines search instead of planes fitting.
We also developed a method to handle data from sites of large size, scanned in high definition, where conventional measuring methods are not appropriate. Moreover, such 3D data offer many possibilities without adding fieldwork (redeployment assumptions, building simulation, and understanding of the entire geometry).

Keywords: Plan extraction, rehabilitation project, primitive fitting, 3D scanner