A GENERIC 3D MODEL FOR AUTOMATED BUILDING ROOF RECONSTRUCTION
S. Scholze“*, T. Moons®°, L. Van Gool“®
“ETH Zurich, Computer Vision Laboratory (BIWI), Zurich, Switzerland
bKatholieke Universiteit Leuven, Center for Processing Speech and Images (ESAT / PSI), Leuven, Belgium
cKatholieke Universiteit Brussel, Group of Exact Sciences, Brussel, Belgium
Working Group V/2
KEY WORDS: Modelling, Reconstruction, Building, Three-dimensional, Aerial, Vision
ABSTRACT
This paper investigates into model-based reconstruction of complex polyhedral building roofs. A roof is modelled as a
structured ensemble of planar polygonal faces. The modelling is done in two different regimes. One focuses on geometry,
whereas the other is ruled by semantics. Inside the geometric regime, which is the primary topic of this paper, 3D line
segments are grouped into planes and further into faces using a Bayesian analysis. In the second regime, the preliminary
geometric models are subject to a semantic interpretation. It is shown, how the knowledge gained in this step can be
used to infer missing parts of the roof model (by invoking the geometric regime once more) and to adjust the overall roof
topology. Several successfully reconstructed complex roof structures corroborate the potential of the approach.
1 INTRODUCTION
Three-dimensional, multiview modelling is a powerful ap-
proach for automated scene interpretation and reconstruc-
tion. Although good results have been obtained from no
other information than image content, applications like 3D
roof reconstruction for buildings from aerial images still
pose a serious challenge. There is often little texture for
correspondence search to work well, and essential features
like straight lines often go undetected. On the other hand,
the class of typical roof shapes is sufficiently constrained
as to supply strong prior knowledge to the system. Deal-
ing with complex man-made objects such as building roofs
in urban areas requires a generic and expressive model.
Finding the right balance between flexibility and predic-
tive power of the model is of key importance to the entire
reconstruction. Although considerable research has been
conducted on model-based building reconstruction, less ef-
fort has been invested in the development of the models
themselves. This paper proposes a parametric model which
is applicable to a broad range of roof types, together with
probabilistic and semantic reasoning procedures to instan-
tiate and finetune the model.
1.1 Previous Work
Despite the long lasting demand for cheap, reliable and
up-to-date 3D city models for technical and environmental
planning tasks, automated building reconstruction is still
an only partially solved task. An excellent overview of im-
portant developments and state-of-the-art in the field can
be found in the Proceedings of the Ascona Workshops on
Automatic Extraction of Man-Made Objects from Aerial
and Space Images (Grün et al., 1995, Grün et al., 1997,
Baltsavias et al., 2001). Some previous approaches, able
to reconstruct complex buildings with minimal user inter-
action include (Henricsson, 1998, Moons et al., 1998, Bail-
lard and Zisserman, 2000, Fischer et al., 1998).
The roof models used in the literature range from very
generic ones (polyhedral structures) up to very specific ones
(parametrized models of different building types). For an
overview see (Mayer, 1999). For handling uncertainty and
imprecision of the input data, some building reconstruction
systems already have a probabilistic underpinning (Cord et
al., 2001, Kulschewski, 1997, Heuel and Fórstner, 2001).
1.20 A Geometric Model in a Probabilistic Setup
In this paper, a probabilistic formulation for geometric build-
ing reconstruction is proposed. Its key component is a
generic and expressive roof model. As a novelty, various
constraints can be explicitly imposed onto the model. The
conjunction of robust geometric reasoning in 3D space to-
gether with a semantic interpretation allows to reconstruct
complex building roofs completely and with correct topol-
ogy. Figure 1 gives an overview of the presented method-
ology.
The organisation of this paper is as follows. A detailed de-
scription of the proposed roof model is presented in Sec-
tion 2. Section 3 mainly states the processing steps from
the image data up to reconstructed 3D line segments. Ad-
ditionally, the notion of semantic labels and the used test
dataset are introduced here. In Section 4, the steps lead-
ing to a preliminary geometric roof model are described.
A summary on how a semantic interpretation of this geo-
metric model is obtained and used to complete and refine
the roof model is given in Section 5. Finally, results of the
approach are presented in Sect. 6.
2 GEOMETRIC ROOF MODEL
Guided by the observation that the broad majority of build-
ing roofs (at least in the western world) consist of planar
surfaces, and that both, internal roof boundaries and the
outline of the entire roof are delineated by straight lines,
a roof model is proposed, which describes a roof as en-
semble of planar polygonal faces (patches). It suffices to
distinguish between triangular and quadrangular patches,
since more complex patch shapes (e.g. L-shapes) are ob-
tained by patch composition. Theoretically, it would suf-
-204—
Figure
systen
ments
some
prelin
3D sp
bellin;
the m
(d) wl
tion. €
overal
fice tc
rangu
cludir
The 3.
be ref
eratel
vision
roof,
and w
2.1
A roc
space
tions,
metri.
of qu
syste1
ordin:
and a
lation
3(a,b)
rangu
or
tion a
ing.
test d
corpc
mode
In pa
tion E
Pi.
repre
