The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008 
1169 
some areas the 3K DSM from 2007 lays above or under the 
DSM from 2006 and 2003. Smaller variations in the surface are 
mainly caused by the seasonal change of the vegetation. 
Height difference 
■■ 30 and more 
mm 20 tc 
m io to 19 
■ HO 10 
rno 
M -1 to-9 
! I-t0 to-'9 
H -20 ta -29 
-3C an« nere 
200 m 
Figure 10 Automatic determination of building heights from 
3KDSM 
Figure 8 Difference of DSM between the years 2006 and 2007. 
Height [m] 
Profile distance [m] 
Figure 9 Profile through DSMs of the years 2003, 2006, and 
2007. 
4.2 Determination of building heights (urban area) 
The automatic determination of building heights could be useful 
e.g. for risk analysis before disasters or damage analysis after 
disasters. In this study, building heights are automatically 
derived from the 3K DSM supported from 3K orthoimages. A 
dynamic threshold operator is applied to the 3K DSM, which 
detects building areas. The orthoimage is used to filter out 
vegetation, which is falsely detected as building. In the 
remaining regions, the maximal difference between 3K DSM 
and the smoothed DSM is determined as building height. 
Figure 10 shows the result of the automatic building height 
algorithm in the center of Munich. 
4.3 Monitoring of urban areas (3D change detection) 
3D change detection in urban area is a quite helpful tool for the 
automatic detection of building modifications, new buildings or 
destroyed buildings e.g. after earthquakes. 
In the airborne case and in particular for high resolution images, 
3D change detection has some big advantages, e.g. a one-to-one 
relation between the same pixels from two different dates is 
prerequisite and could not be obtained in the urban area without 
using the 3D information. 
Nevertheless, 3D change detection is a quite complex task. 
Problems arise not only when comparing two DSM from 
different sensors, like LIDAR and optical sensor. Also the 
comparison of two DSMs from the same sensor is quite 
complex, as often different faces of the buildings are exposed to 
the sensor, which causes variations and errors in the DSM. 
First experiences with 3D change detection using DSM of 3K 
camera system were made with image data from 30 th April and 
17 th June 2007 over Munich. A LIDAR reference DEM from 
the same area was acquired earlier by the Bavarian map 
authority. Flight height at both dates was 2000m a.G. Three 
consecutive images were taken for DSM generation based on a 
height base ratio of 1:17 at 30 th April and 1:4 at 17 th June 2007. 
Figure 11 shows the difference between the 3K DSM from 17 Ih 
June and the LIDAR DEM. To isolate building modifications or 
damages from other disturbances, morphological operations and 
thresholds (not discussed here) must be applied to the difference 
DSM. Detected 3D building changes between the acquisition 
dates of the DEMs are marked as red regions, e.g. the newly 
built Jewish Synagogue (left) and the “Schrannenhalle” (right) 
in the center of Munich. 
4.4 Monitoring of infrastructure 
Up-to-date information about the state of infrastructure, in 
particular about the roads, is important information for BOS 
after disasters. Mostly, this information can be derived directly 
from the images, but in some cases the 3D information could be 
helpful. In combination with a road data base, automatic tools 
could detect e.g. bridges or bridge damages using the 
georeferenced image and the 3D information from the DSM. 
Figure 12 shows an example where the road database provides