In: Stilla U, Rottensteiner F, Paparoditis N (Eds) CMRT09. IAPRS, Voi. XXXVIII, Part 3/W4 — Paris, France, 3-4 September, 2009 
SEMI-AUTOMATIC CITY MODEL EXTRACTION FROM TRI-STEREOSCOPIC VHR 
SATELLITE IMAGERY 
F. Tack a- *, R. Goossens a , G. Buyuksalih b 
d Dept, of Geography, Ghent University, Krijgslaan 281, 9000 Ghent, Belgium - (f.tack, rudi.goossens)@ugent.be 
b IMP-Bimtas, 34430 Beyoglu, Istanbul, Turkey - gbuyuksalih@yahoo.com 
KEY WORDS: Photogrammetry, City DSM generation, Tri-stereoscopy, Ikonos, Urban 
ABSTRACT: 
In this paper a methodology and results of semi-automatic city DSM extraction from an Ikonos triplet, is introduced. Built-up areas 
are known as being complex for photogrammetric purposes, mainly because of the steep changes in elevation caused by buildings 
and urban features. To make surface model extraction more robust and to cope with the specific problems of height displacement, 
concealed areas and shadow, a multi-image based approach is followed. For the VHR tri-stereoscopic study an area extending from 
the centre of Istanbul to the urban fringe is chosen. Research concentrates on the development of methods to optimize the extraction 
of a surface model from the bundled Ikonos triplet over an urban area, without manual plotting of buildings. Optimal methods need 
to be found to improve the radiometry and geometric alignment of the multi-temporal imagery, to optimize the semi-automatical 
derivation of DSMs from an urban environment and to enhance the quality of the resulting surface model and especially to reduce 
smoothing effects by applying spatial filters. 
1. INTRODUCTION 
The high level of detail and geometric accuracy of very high 
resolution satellite data such as Ikonos imagery, has made this 
kind of imagery suitable for DSM generation at feature level of 
urban environments. Due to the photogrammetric complexity of 
urban areas, quite some research is done to cope with the 
specific problems of urban surface model generation from 
standard stereopairs. As a multi-image based approach can 
make the 3D modelling more robust, a methodology and results 
of semi-automatic DSM production from an Ikonos triplet over 
an urban area, is highlighted in this paper. From a theoretical 
point of view the redundancy of a third image should lead to a 
more reliable photogrammetric processing. Only a few 
investigations have been published dealing with the concerning 
subject. Research published in (Baltsavias et al., 2006) and 
(Raggam, 2006) can be referred to. 
Research is conducted within the framework of the MAMUD 
project (Measuring And Modelling of Urban Dynamics) funded 
by the STEREO (Support to The Exploitation and Research of 
Earth Observation data) program of Belgian Science Policy. 
The objectives of the MAMUD research project is to investigate 
the possibilities of earth observation for a better monitoring, 
modelling and understanding of urban growth and land-use 
change. Urban change processes are affecting the human and 
natural environment in a not unimportant way. This enlarges the 
need for more effective urban management approaches based on 
sustainable development. A sustainable urban management 
needs sufficiently detailed and reliable base information on the 
urban environment and its dynamics. Satellite imagery has 
proven to be an important data source to monitor and describe 
urban areas and its changes. Hereby, detailed information on the 
vertical structure is vital to label urban features, to describe 
urban morphology and to generate spatial metrics. If the 
subsequent approach is proved to be successful, it will increase 
the flexibility of producing semi-automatic 3D city models from 
high resolution satellite imagery. 
The complexity of an urban environment for photogrammetric 
purposes will be highlighted in section 2. In section 3 the image 
dataset and work area will be outlined. The different phases of 
the photogrammetric processing of the Ikonos triplet are stated 
in section 4. In following section, spatial filtering is applied on 
the height values of the surface model to improve the quality 
and reduce smoothing effects. Geometric accuracy analysis is 
discussed in section 6. Finally, in section 7 experiences and 
conclusions are summarized. 
2. COMPLEXITY OF URBAN AREAS 
A Digital Surface Model is a digital representation of the terrain 
and topographic object height in a grid structure. Interpolation 
of the discrete height values is needed to approximate the 
continuity of the ground surface. Urban environments are 
experienced as complex for 3D modelling purposes because of 
the steep changes in elevation and the discrepancy between the 
smoothness of the ground surface and abrupt discontinuities 
caused by buildings and other urban features. Without manual 
plotting or spatial filter techniques it is difficult to reconstruct 
vertical walls out of VHR satellite imagery. An interpolation 
technique creates a smoothed surface and causes individual 
buildings will have a shape of a bell instead of the rectangular 
geometry (Jacobsen, 2006). A second consequence of steep 
changes in elevation is the occurrence of shadow and concealed 
areas. Due to the convergent viewing angle of VHR sensors like 
Ikonos, terrain features with certain height above the surface are 
geometrically displaced in the imagery, leading to 
dissimilarities between the stereo images. 
By this distortion of its true position, parts of the ground surface 
can be hidden in the satellite image. These are so-called 
occluded areas. Shadow areas, which have poor contrast, and 
Corresponding author