ZA, 9-11 Nov. 1999 
esponding image section 
ace is rectified applying a 
apped to the corresponding 
t of this process is visualized 
  
  
International Archives of Photogrammetry and Hernote Sensing, Vol. 32, Part 3W14, La Jolla, CA, 9-11 Nov. 1999 
Figure 9: Visualization of buildings after texture mapping of 
faccades. 
5 ALTERNATIVE DATA SOURCES 
Up to now airborne laser scanning was applied to provide the 
geometric information for building reconstruction, whereas the 
image texture was extracted from aerial or terrestrial imagery. 
Alternatively, the aerial images can also be applied for height 
data acquisition. This would be advantageous since the same 
sensor - an airborne camera - can be used to provide the 
required geometric and radiometric information. Currently, the 
application of aerial images for DSM generation is mainly 
hindered by the problems of automatic image matching at 
occlusions and height discontinuities, which especially occur in 
built-up areas. These problems can be avoided at least to a 
certain extend if the applied imagery is captured at a relatively 
large scale by a normal angle camera. 
  
Figure 10: DSM from airborne laser scanning 
  
Figure 11: DSM from stereo image matching 
An example for the DSM quality, which can be obtained in 
urban areas is given in Figure 11 and Figure 10. Figure 10 
shows a grayvalue representation of height data acquired by 
laser scanning, Figure 11 gives the corresponding result 
produced by stereo image matching. In order to generate this 
DSM the stereo image pair was captured at an image scale of 
1:5000 using a aerial camera with focal length of 305 mm. The 
images were scanned with 15 pum resulting in a ground pixel 
size of 7.5 cm for each pixel. For DSM generation the standard 
software tool MATCH-T was applied (Krystek, 1991). In this 
example the results of airborne laser scanning and stereo image 
matching can be compared very well since both DSM are 
collected with a grid with of 1 m. Whereas for the DSM from 
image matching smooth transitions between roof and terrain 
surface are visible for some areas, the breaklines are defined 
more sharply in the DSM from laser scanning. Additionally, 
details like the small tower are only present in the laser DSM. 
At the moment the direct height measurement by airborne laser 
scanners provides DSM data of higher and more homogeneous 
quality especially in urban areas. Nevertheless a lot of effort has 
been spent on software development to overcome the 
deficiencies of stereo image matching in build-up areas. 
Examples are the integration of knowledge on predefined 
breaklines or the use of multiple images (Schliiter, 1998). An 
additional improvement of the derived DSM data can be 
expected by the increasing quality of original stereo image data 
if digital aerial cameras are applied (Renouard and Lehmann, 
1999). For these reasons, the future development of both 
techniques for DSM generation concerning accuracy, reliability 
but also cost effectiveness has to be observed carefully. 
6 CONCLUSION 
Within the paper an approach for the generation of 3D urban 
models based on height data from laser scanning and existing 
ground plans has been presented. The process enables a fully 
automatic reconstruction of the buildings as well as an 
interactive refinement of the results provided by the automatic 
procedure. This is a main advantage of the presented approach, 
since a fully automatic reconstruction is feasible for large areas, 
whereas the refinement can be restricted to important parts and 
can therefore be performed only on demand. Another advantage 
of using a DSM for virtual landscape generation is the good 
aptitude of height data for change detection. A change in height 
must result from a change in surface geometry, whereas a gray