XIX-B3, 2012
as, U. and Reinartz,
from airborne digital
ngress of the ISPRS,
he Photogrammetry,
ences.
'ametric house mod-
ISPRS Workshop on
land.
2002. Building ex-
A Research Report.
D., Jacobsen, K. and
ality analysis of dem
1 optical stereo satel-
ISPRS. ISPRS Sym-
Canada, International
nsing and Spatial In-
id Kubik, K., 2007.
ser scanner data and
ation and sensitivity
and Remote Sensing
building detection in
nference on Control,
/ 2008, Hanoi, Viet-
babilistic framework
ages. IEEE Transac-
). —.
] Reinartz, P., 2010.
sing automated com-
ion VII Symposium,
ind Ehlers, M., 2012.
seneration of six dif-
yrs. IEEE Journal of
ns and Remote Sens-
. Image processing,
tions, Pacific Grove,
upt, M., 2010. Anef-
print extraction from
Photogrammetry and
2008. Classification
ita and digital eleva-
and Remote Sensing
sch, T. and Dech, S
on using multisensor
ernational Journal of
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B3, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
PERFORMANCE ASSESSMENT OF FULLY AUTOMATIC THREE-DIMENSIONAL
CITY MODEL RECONSTRUCTION METHODS
Beril Sirmacek, Hossein Arefi, Thomas Krauss
German Aerospace Center (DLR), Earth Observation Center (EOC)
PO Box 1116, 82234, Wessling, Germany
(Beril.Sirmacek, Hossein.Arefi, Thomas.Krauss) G dlr.de
Commission III/4
KEY WORDS: Building, Modelling, City, DBM/DTM
ABSTRACT:
Three-dimensional urban region representations can be used for detailed urban monitoring, change and damage detection purposes. In
order to obtain three-dimensional representation, one of the easiest and cheapest way is to use Digital Surface Models (DSMs) which
are generated from very high resolution stereo satellite images using stereovision techniques. Unfortunately after applying the DSM
generation process, we cannot directly obtain three-dimensional urban region representation. In the DSM which is generated using only
one stereo image pair, generally noise, matching errors, and uncertainty on building wall locations are very high. These undesirable
effects prevents a DSM to provide a realistic three-dimensional city representation. Therefore, some automatic techniques should be
applied to obtain three-dimensional city models using DSMs as input. In order to solve the existing problems in this field, herein we
introduce two automated approaches based on usage of DSMs as input. The first method depends on using of a 3D active shape model
for building shape extraction and 3D reconstruction, the second approach is based on an approximation of prismatic models to DSMs.
Our experimental results on images and DSMs of Tunis city which are obtained from WorldView-2 satellite indicate possible usage of
the proposed algorithms to obtain three-dimensional city representations automatically.
1 INTRODUCTION
Three-dimensional representations of complex environments ob-
tained a lot of interest for various applications with the develop-
ment of satellite sensor technology. In order to obtain 3D repre-
sentations of cities, one of the most practical method is to gen-
erate Digital Surface Models (DSMs) using very high resolution
satellite images which are taken from two or more viewing an-
gles. Unfortunately, due to occlusions, matching errors, and ap-
plied interpolation techniques these DSMs do not represent 3D
city models directly. Automatically generated DSMs do not con-
tain sharp wall and rooftop representations. Therefore, in order
to obtain real 3D city representations, advanced methods should
be applied to DSMs.
In the related literature, Brunn and Weidner (Brunn and Weidner,
1997) used surface normals on DSM to discriminate buildings
and vegetation. After detecting buildings, they measured the ge-
ometry of rooftops using surface normals and they interpolated
polyhedral building descriptions to these structures. Fradkin et
al. (Fradkin et al., 1999) proposed a segmentation based method
to reconstruct three-dimensional models of dense urban areas. To
this end, they used very high resolution color aerial images and
disparity maps. Canu et al. (Canu et al., 1996) used a high res-
olution DSM, which is obtained by stereo matching techniques,
In order to reconstruct three-dimensional buildings. First, they
segmented the DSM into homogeneous regions. Then, they in-
terpolated flat surfaces on these regions. Ortner et al. (Ortner
et al., 2002) used a point process (Jacobsen, 2005) to model ur-
ban areas. They represented urban areas as interacting particles
Where each particle stands for an urban object. Preknowledge
about building shapes is used to model these particles. Arefi et
al. (Arefi et al., 2008) extracted above-ground objects from LI-
DAR data. Then, three-dimensional buildings are reconstructed
by hierarchical fitting of minimum boundary rectangles (MBR)
and a RANSAC based straight line fitting algorithm. Tournaire
et al. (Tournaire et al., 2010), developed a stochastic geome-
try based algorithm to detect building footprints from DSM data
which have less than 1 m resolution. They tried to fit rectangles
on the buildings using an energy function and prior knowledge
about buildings. To minimize the energy function, they used a
Reversible Jump Monte Carlo Markov Chain (RJMCMC) sam-
pler coupled with a simulated annealing algorithm which leads
to an optimal configuration of objects. Maas (Maas, 1999) used
maximum slope values in order to determine best fitting rooftype
shapes to generate three-dimensional building models. Valero et
al. (Valero et al., 2008) developed a feature extraction and classi-
fication based method to classify building roofs into two classes
as flat-roof and gable-roof. They estimated ridge-line positions
which are based on skeletons of groundfloor plans. They pro-
vided the difference between the average roof outline height and
the average ridge-line height as first feature, and the norm of
the orthorectified image gradient as second feature for the sup-
port vector machine (SVM) classifier. In all introduced stud-
ies, good results are achieved generally using very high resolu-
tion (better than 1 m spatial resolution) DSMs which are gener-
ally generated from airborne images or LASER scan data. How-
ever, enhancement of buildings in low resolution urban DSM data
which are generated from satellite images is still an open research
problem. In order to bring an automated solution to this prob-
lem, in previous work Sirmacek et al. proposed a novel tech-
nique for obtaining three-dimensional city representations by ap-
plying a building shape and rooftop-type detection approach to
DSMs (Sirmacek et al., 2011). They started by applying local
thresholding to raw DSMs in order to extract high urban objects
which can indicate building locations. They extracted building
shapes from regions which are obtained from thresholding result
by using a binary active shape growing algorithm that they pro-
posed which depends on growing rectangular shapes in elongated
segments which are detected in binary mask which is obtained
by thresholding the DSM. After obtaining building shapes, they
generated three-dimensional models by understanding building
