Detection, Modeling, and Classification of Moldings for Automated Reverse Engineering of Buildings from 3D Data

Laser scanner data is increasingly being used for the detailed reverse engineering of buildings. This process is currently primarily manual, but recent research has shown that basic structures, such as walls, ceilings, floors, doorways, and windows, can be detected and modeled automatically. Building on this previous research, we focus on the modeling of those linear moldings that typically surround doorways, windows, and divide ceilings from walls and walls from floors. These structures may be secondary and merely ornamental, but many projects nevertheless require that they be modeled. Moldings can be difficult to model manually owing to missing data caused by occlusions or to the ambiguity caused by low data density. Our molding modeling approach consists of two steps: 1) estimating the path of the molding; and 2) estimating the shape of the molding profile. In the first step, we iteratively update the molding’s line of extrusion by optimizing the similarity of cross-sections sampled along the path, thereby compensating for imperfections in the initial orientation estimate. In the second step, a unified profile is extracted using data from the entire length of the molding, which allows for partial missing data from occlusions. The profile is then characterized by a specific shape descriptor. Finally, a KNN algorithm classifies the molding into a database which has been constructed with profiles originating from various molding manufacturers. We demonstrate the method using real 3D laser scanner data of various types of moldings, both simple and complex.