GENERATION AND ANALYSIS OF DIGITAL SURFACE MODELS (DSMS) IN URBAN 
AREAS BASED ON DMC-IMAGES 
A. Alobeid, K. Jacobsen 
Institute of Photogrammetry and Geoinformation , Leibniz University Hannover 
Nienburger Str. 1, D-30167 Hannover, Germany - 
(Alobeid, Jacobsen)@ ipi.uni-hannover.de 
Commission IV, WG IV/9 
KEY WORDS: Digital, Matching, DEM/DTM, Aerial, Accuracy, Quality 
ABSTRACT: 
DSMs are a geometric description of the height of the visible terrain surface, including elements like buildings and trees. DSMs 
became an important source for scene analysis and for 3D feature detection and reconstruction. The digital metric camera (DMC) 
offers excellent image quality and can be used for automated DSM generation, with high accuracy. By using least squares matching 
as method to find corresponding points in two overlapping images for three dimensional reconstructions, the corresponding points 
potentially can be determined with sub-pixel accuracy. The investigation of the influence of different parameters on generating 
DSMs from DMC images is shown like the limitation of automatic DSM generation in urban areas. The aim of our investigation is, 
whether the selected parameters for image matching, optimized in few test areas, can be used for the whole study area. The 
generated DSM was checked against a reference DSM based on manual measurement. The comparison shows that the accuracy 
mainly depends on terrain slope and surface structure. Furthermore, satisfying DSMs strongly depend upon the quality of the DMC- 
images.A standard deviation of the DSM based on DMC-images in whole urban study area of 1.0 up to 1.3 GSD is possible, 
corresponding to 0.3 up to 0.4 GSD for the x-parallax. 
1. INTRODUCTION 
Automatic 3D city models and its update got an increasing 
interest in recent years, requiring the development of automatic 
methods for acquisition of DSMs. 3D information in urban and 
suburban areas are very useful for many applications, such as: 
- Change detection 
- GIS database updating 
- Urban planning 
- 3D feature extraction 
In principle stereo images are sufficient for 3D feature 
extraction for objects such as buildings. However due to the 
complexity of details, many problems were found. Grey values 
are influenced by the object geometry but also by many factors 
such as shadows, reflections, saturation, lack of texture, noise 
and others. 
It is very difficult to separate important information from 
irrelevant details in generating DSMs in build-up areas. In spite 
of high degree of automation in several commercial programs, 
used for DSM generation, a significant effort is still required for 
accuracy improvement, especially in the difficult urban and 
forest areas. The aim of our investigation was to investigate the 
influence of different parameters on generating DSMs based on 
Z/I Imaging DMC-images in urban areas. In addition the use of 
the selected parameters for other areas has been investigated. 
2.AVAILABLE DATA AND STUDY AREA 
resolution panchromatic sub images, mosaicked together 
(see figure 1). The convergent camera arrangement guarantees a 
very uniform image quality up to the image comers, no lower 
resolution in the comers could be detected by edge analysis. 
This is different for digital cameras based on nadir view for all 
sub-cameras (Jacobsen 2008). 
Figure. 1 Configuration of panchromatic DMC sub-cameras 
The images are available with a ground sampling distance 
(GSD) of 9.4cm having 60% endlap and 75% side lap. The 
horizontal coordinate components of well defined object points 
have a standard deviation of approximately 2-3cm and the 
vertical component of approximately 7cm. 
The investigations have been made in the area of Frederikstad, 
located close to Oslo, Norway. The Digital Metric Camera 
(DMC) images are composed of 4 slightly convergent high