PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
June 2015
481
Object-Based Building Change Detection from a
Single Multispectral Image and Pre-Existing
Geospatial Information
Georgia Doxani, Konstantinos Karantzalos, and Maria Tsakiri-Strati
Abstract
Multispectral images of very high spatial resolution and vector data
from geospatial databases, such as cadastral maps, are among
the cost-effective and broadly available geodata in urban environ-
ments. Therefore, we aim to address building change detection
based on pre-existing building footprint information and a single
very high resolution multispectral image. An object-based classifi-
cation methodology was developed that employs advanced scale-
space filtering, unsupervised clustering, and knowledge-based
classification. The developed framework effectively integrates prior
vector data and multispectral observations, through incorporating
the prior knowledge into the training process and defining the
proper object-based classification rules. The methodology success-
fully identified important building changes, which were validated
by employing the vector information of a building geodatabase and
a QuickBird image acquired in 2003 and 2007, respectively, over
urban regions in the city of Thessaloniki, Greece. The performed
quantitative and qualitative evaluation indicated that the proposed
analysis framework can detect the new buildings with high accura-
cy rates and, to a lesser degree, their exact shape and size.
Introduction
The development of operational procedures for the automated
or semi-automated map and geodatabase updating has been
investigated thoroughly during the last decade. Although, the
traditional analytical/manual ground survey and photogram-
metric approaches are providing detailed and accurate change
detection products, the recently available earth observation
datasets of very high spatial and spectral resolution, question
their time- and cost- effectiveness. In particular, a number of
sophisticated algorithms and techniques for change detection
have been proposed providing a cost-effective solution for
detecting spatial patterns in a repeatable way (Holland
et al
.,
2008; Leignel
et al
., 2010; Matikainen
et al
., 2012; Mallinis
et
al
., 2014; Karantzalos, 2015).
Certain studies for building change detection employ 3D
information derived mainly from laser data or stereo image
pairs to facilitate the distinction of buildings from the other
ground objects with similar spectral responses. The main con-
cept of those approaches is the creation of a mask with all the
above-ground objects based on height information and then the
refinement of this mask by utilizing the spectral information
of satellite or aerial imagery. Such an approach is proposed by
Matikainen
et al
. (2010) towards the automatic building ex-
traction for map updating purposes. In their study, the integra-
tion of aerial color images, airborne laser scanning data, and an
existing building map were employed for the detection of the
new buildings in the study area. A Digital Surface Model (
DSM
),
which was derived from the laser data, was used to define the
above ground objects, such as buildings and trees. Based on a
range of attributes and some training data, obtained from the
existing building map, a classification tree was created auto-
matically for distinguishing the buildings from the trees. The
approach was efficient at detecting most of the buildings larger
than 60 m
2
, and the errors observed were mainly due to small
buildings that were difficult to be identified.
Building change detection has been also proposed through
the integration of pre-existing vector information, multiple
satellite images, and a
DSM
derived from those images (Cham-
pion
et al
., 2010). The proposed two-step methodology first
carries out the extraction of 2D and 3D primitives from the
multiple satellite images and the
DSM
. The extracted primitives
were then matched with the primitives of an existing building
database to define three ‘coarse’ building classes (demolished,
modified, and unchanged) that could then be used to undertake
a partial update of the database. In the second step, a mask,
which was extracted from the
DSM
and contained the above-
ground objects, in combination with a computed Digital Terrain
Model (
DTM
) were integrated to refine the initial classification.
In a similar manner, building extraction in suburban areas
was also the objective of Grigillo
et al
. (2012). Initially, all possi-
ble building changes were detected based on a normalized
DSM
.
The final changes were derived by the application of a process-
ing scheme on multispectral data, including non-linear diffu-
sion filtering, unsupervised classification, color segmentation,
and region growing. The quantitative evaluation indicated that
the majority (83.2 percent) of the buildings had been detected,
while the overall resulting quality was around 50 percent.
Model-based approaches have also been employed in
building change detection studies. Prior knowledge and infor-
mation are coupled with the observations from imagery, while
object-based information, i.e., color, texture, shape, size, and
topological information, facilitate the extraction procedure
(Baltsavias, 2004; Mayer, 2008). To this end, prior knowledge
has been employed by many change detection studies (Bail-
Georgia Doxani is with the European Space Agency/ESRIN,
Via Galileo Galilei, 00044, Rome, Italy; and the Department of
Cadastre, Photogrammetry and Cartography, Faculty of Rural
and Surveying Engineering, Aristotle University of Thessalon-
iki, University Campus, 54124, Thessaloniki, Greece (Georgia.
).
Konstantinos Karantzalos is with the Department of Cadas-
tre, Photogrammetry and Cartography, Faculty of Rural and
Surveying Engineering, Aristotle University of Thessaloniki,
University Campus, 54124, Thessaloniki, Greece.
Maria Tsakiri-Strati is with the Remote Sensing Laboratory,
School of Rural and Surveying Engineering, National Tech-
nical University of Athens, Heroon Polytechniou 9, Zografos,
15780, Athens, Greece.
Photogrammetric Engineering & Remote Sensing
Vol. 81, No. 6, June 2015, pp. 481–489.
0099-1112/15/481–489
© 2015 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.81.6.481
