inbul 2004 
ipplication 
d used for 
ment tasks 
| can have 
the degree 
> changes. 
ding states 
üssing (or 
'endent on 
ub-divided 
following 
exclusively 
yses those 
1 approach 
ctection is 
ed 
hat do not 
er date to 
1ot-altered. 
d in more 
| compared 
1sor is less 
; could be 
:viations is 
en weakly 
ects do not 
missing in 
asurements 
sations of) 
sis of laser 
e regarded 
ent data. It 
yartly resp. 
ifferentiate 
ack ground 
ble enough 
the roof in 
ntly higher 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
  
Figure 3: Aerial photography 
of a dark roof 
Figure 4: Same scenery in 
grey-value coded first 
pulse laser scanning 
DSM (TopoSys I) 
  
  
Figure 6: ... and last pulse 
laser scanning DSM 
( TopoS ys II) 
Figure 5: Same scenery in last 
pulse laser scanning DSM 
(TopoSys I) 
Buildings containing large parts without data can not be 
regarded further in a change detection analysis as their state is 
not observable at one date. 
Therefore, in a first step all building segments derived from 
laser scanning data were consecutively examined with regard to 
included no-data pixels in the according laser scanning DSM 
and were filtered out if a sufficient number of such pixels 
occurred. Here, the threshold was set to 4, assuming that no- 
data pixels may occur at elements on a roof. ¢.g. chimneys, but 
it can not be tolerated if they cover large parts of a building; on 
the other hand small buildings can consist of only 10 pixel (e.g. 
2m x 5m ground plan). therefore, this value should not be 
exceeded. 
4.2 New and demolished buildings 
In a next step, the overlaps of the extracted building areas are 
examined. Based on the segments received from the DSM of the 
earlier date ( DSM(t;) ), all pixels of a segment are analysed if 
they are part of a segment derived from the newer DSM 
( DSM(t,) ). too. Only if a segment of t, is found which has a 
certain percentage (e.g. 2596) of its ground plan area 
overlapping with segments of t;, it 1s kept for a further analysis. 
Otherwise, at all pixels the differences in between the both 
DSMs are checked. If a significant number of pixels - here the 
inverse percentage of the overlapping threshold is used (75%) - 
exceed a deviation tolerance value, which is introduced due to 
measurement accuracy and systematic deviations on slant 
planes, the segment is assumed to represent a demolished or 
teared off building. Otherwise the segment is also regarded for 
further analysis. The same procedure is carried out also on the 
segmentation result of the newer DSM and accordingly new 
buildings are found. 
In the Figure 9 an example of the obtained results is shown. For 
an easier understanding the scenery is shown in two aerial 
images (Figure 7 and Figure 8), which were acquired 
approximately at the same time when the laser scanning flights 
took place, and, in Figure 9, the segments extracted from DSM 
acquired in 2002. 
  
Figure 7: Acrial Photography captured in 1997 
same scenery photographed in 2001 
Figure 8: ...