International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
scanning was presented. In a first step, the laser data has been
segmented and classified. In the second step, the laser data
segments of buildings have been matched against the
building objects of a topographical database.
With respect to the classification results several conclusions
can be drawn. In general, laser data can be classified
relatively reliable. However, to really allow fully automatic
change detection and to ensure a low percentage of incorrect
change detections further improvements are required. The
largest problem in this respect is caused by vegetation
adjacent to buildings. If this vegetation is considered as an
extension of a building, this error will generate an incorrect
signal for the need of a database update and thus require extra
operator time. In this research we used average point
distances of 1.2-1.4 m. Higher point densities may allow
better classifications.
For the classification experiments in this paper usage was
made of both roughness and colour information. Colour
information appeared to be a useful addition. The
classification accuracy of buildings was improved by 3%.
The additional value of colour information may, however,
vary from project to project and depend on the season and the
colours of the roofs. Classification results should further
improve with the additional usage of multiple pulse data.
In the change detection experiment all newly constructed
buildings were detected reliably. Differences between laser
data segments and database objects caused by generalisation
or data noise could effectively be handled by mathematical
morphology. More challenging is the implementation of the
object selection rules as laid down in the mapping catalogue.
In the case of the TOP1Ovector database, the definition of
what to map was sometimes vague and often required a
certain amount of scene interpretation. For the purpose of
automatic change detection the rules of the mapping
catalogues need to be defined more precisely and preferably
avoid the usage of definitions which require semantic
modelling for the interpretation of the scenery.
In the performed experiments several errors caused by the
mapping process of the topographical database have been
found. This showed that automatic change detection can
already now be a useful tool for quality control despite
limitations in the classification accuracy and the
interpretation of mapping rules.
ACKNOWLEDGEMENTS
The laser data for this study was provided by Terralmaging
B.V. The TOPlOvector data was provided by the
Topografische Dienst Nederland. The authors would like to
thank both organisations for their support of this research.
REFERENCES
Kessels, P., 2003. Automatic change detection of buildings in
the TOP10vector map using airborne laser scanner data (in
Dutch). M.Sc.-thesis Delft University of Technology.
Maas, H.-G., 1999. The Potential of Height Texture
Measures for the Segmentation of Airborne Laser Scanner
Data. In: Proceedings of the Fourth International Airborne
Remote Sensing Conference, Ottawa, Ontario, Canada, pp.
154-161.
Matikainen, L., Hyyppä, J. Hyyppä, H., 2003. Automatic
Detection of Buildings from Laser Scanner Data for Map
Updating. In: International Archives of Photogrammetry,
Remote Sensing and Spatial Information Sciences, Dresden,
Germany, vol. 34, part 3/W13, pp. 218-224.
Murakami, H., Nakagawa, K., Hasegawa, H., Shibata, T. and
Iwanami, E., 1999. Change Detection of Buildings using an
Airborne Laser Scanner. /SPRS Journal of Photogrammetry
and Remote Sensing 54 (2-3): 148-152.
Murakami, H., Nakagawa, K., Shibata, T. and Iwanami, E.,
1998. Potential of an Airborne Laser Scanner System for
Change Detection of Urban Features and Orthoimage
Development. In: /nternational Archives of Photogrammetry
and Remote Sensing, Stuttgart, Germany, vol. 32, part 4, pp.
422-427.
Oude Elberink, S. Mans, H-G. 2000. The Use of
Anisotropic Height Texture Measures for the Segmentation
of Airborne Laser Scanner Data. In: /nternational Archives of
Photogrammetry and Remote Sensing, Amsterdam, The
Netherlands, vol. 33, part B3/2, pp. 678-684.
Sithole, G., Vosselman, G., 2003. Automatic structure
detection in a point cloud of an urban landscape. In:
Proceedings 2" GRSS/ISPRS Joint Workshop on Remote
Sensing and Data Fusion over Urban Areas, URBAN2003,
Berlin, Germany, pp. 67-71.
Sithole, G., Vosselman, G., 2004. Experimental Comparison
of Filter Algorithms for Bare Earth Extraction from Airborne
Laser Scanning Point Clouds. Accepted for publication in the
ISPRS Journal of Photogrammetry and Remote Sensing.
Steinle, E., Oliveira, F.H., Báhr, H.-P., Loch, C., 1999.
Assessment of Laser Scanning Technology for Change
Detection in Buildings. In: Proceedings XVIIth CIPA
Symposium, October 3-6, Olinda, Brazil. Accessed from
http://www-ipf.bau-verm.uni-karlsruhe.de/Projekte/SFB461/
C5/publications/Olinda99.pdf on April 14, 2004.
Voegtle, T., Steinle, E., 2003. On the Quality of Object
Classification and Automated Building Modelling Based on
Laserscanning Data. In: International | Archives of
Photogrammetry, Remote Sensing and Spatial Information
Sciences, Dresden, Germany, vol. 34, part 3/W13, pp. 149-
155.
KEY WC
ABSTRA
The incre
work. In
scanning
urban fea
variable v
data with
the roof I
Results fr
3D city rr
applicatio
building
approach
source. H
the perfor
With the
conventio
LIDAR L
and featu
been sho
3D featur
of data su
Haala, 19
LIDAR d
point, the
display (
from suck
been trie
some lim
and recon
In earlier
algorithm
building
excluding
segmenta
histogram
inference
determine
the secon
footprints
After obt
each buil
Processin
necessary
a refinem
step utiliz
