CIP A 2005 XX International Symposium, 26 September - 01 October, 2005, Torino, Italy 
746 
AUTOMATING THE EXTRACTION OF REVOLUTION OBJECTS FROM SINGLE 
LASER SCANS OF ARCHITECTURAL SCENES 
M. Deveau, G. Letellier, N. Paparoditis 
Institut Géographique National, Laboratoire MATIS, 2, 4 avenue Pasteur 94165 Saint-Mandé Cedex 
matthieu.deveau@ign.fr, guillaumeletellier@free.fr, nicolas.paparoditis@ign.fr 
KEY WORDS: scanner laser, 3D Reconstruction, surface of revolution, profile estimation, range image 
ABSTRACT 
This paper focuses on the reconstruction of surfaces of revolution. We propose a method for estimating those surfaces from point 
clouds. We are more particularly dealing with data coming from one unique laser scanner acquisition; indeed, our approach 
extensively uses the range image topology. It makes originally use of geometric properties of surfaces of revolution. We firstly get an 
approximation of the axis of revolution, thanks to a Hough transform of symmetry axis computed on range image contours. Robust 
and accurate axis estimation refinement of position and orientation is then reached by minimizing the rms error of the profile 
estimate. Image topology is also used to estimate the object’s profile. 3D points are connected following the image topology and the 
nearest neighbouring planes rotating around the axis of revolution. After projection on these planes, lines connecting points allow 
easy profile estimation. This method solves the measuring noise issue which is usually sorted out through moving least squares 
solutions. Once the object of revolution has been reconstructed, we retrieve automatically all occurrences of this pattern in the scene, 
with a scale adaptative matching algorithm. 
1. GENERAL CONTEXT 
The ARCHIPOLIS project, from the MATIS laboratory of the 
Institute G éographique National (IGN), is dealing with high 
scale city model reconstruction. A general strategy is proposed 
combining aerial and terrestrial high resolution stereo imaging 
on streets and fac.ades and dense laser and image acquisition on 
main cultural heritage buildings. Complementarity and 
continuity with aerial building reconstruction methods is also 
studied (Paparoditis et al., 2005). 
In this framwork, detailed reconstruction of cultural heritage 
buildings from laser data and image is the scope of this paper. 
Architectural scenes reconstruction of great heritage value is 
especially important to reach a realistic representation. In this 
context, one goal is to get a compact structured surface repre 
sentation, with a high geometric accuracy. We aim metrological 
applications for architects and surveyors, for restoration purpose 
and archiving. 
As cultural heritage buildings have got more complex structures 
than apartments blocks, photogrammetry leads to long surveys, 
that much longer than architecture is complex. On the other 
side, laser scanners bring quickly a dense geometrical 
information on complex parts. In this context, we are studying a 
methodology using both laser data and image to ease 
segmentation and reconstruction (Deveau et al., 2005). In this 
article, we consider the segmentation part achieved and deal 
with the reconstruction part of/on one specific structure: the 
Surface Of Revolution (SOR), caracterized by its profile and its 
axis of revolution. In this article, which reports our first works 
on this field, only laser data are considered to achieve 
reconstruction while image is only used for texturing. 
1.1 Related works 
Many solutions for the reconstruction of SOR have been pro 
posed and are still studied. Many of them deal with a single 
perspective view, either calibrated (Gr "un, 1975) (Wong et al., 
2002) or uncalibrated (Utcke and Zisserman, 2003) (Colombo et 
al., 2005) and from silhouette’s contours. (Neviata et al., 1995) 
have proposed methods of estimating generalized cylinders 
from multiple images. Other ways of looking for SOR 
reconstruction have followed the growth of range measuring 
systems (Pottmann et al., 2002) (Willis et al., 2001). These 
latter approaches have the drawback to lead to non-linear 
system minimization. 
1.2 Data characterization 
We make use of the range image and the intensity image, the 
latter coming from the backing laser intensity measure and 
belonging to the same image geometry as the range image 
(figure 1.2). Objects we are considering are usually scanned so 
that more than one thousand points lie on their surface. Intensity 
range is spread out on 8 bits. Standard deviation noise on 
distance is calibrated to 2 mm at 10 m (scanning conditions 
used for our experiments). 
Figure 2: High resolution image. 
2. EXTRACTING THE OBJECT SILHOUETTE 
The SOR object is generally bounded by discontinuities which 
are well characterized both in the range image and in the 
intensity image. And because the SOR bounds symmetry 
around axis of revolution is kept in the range image, a first axis 
estimate can be solved. 
We use a Hough transform on contour points coming from a