2004 
  
nated 
were 
luring 
f the 
id of 
tched 
ith an 
) was 
cloud. 
lected 
eneral 
thus 
d to a 
a total 
ed for 
ion of 
CK 
out in 
s were 
ds due 
  
International Archives of the Photogrammetry, Remote Sensing 
to time limitations for the recording. The entire historical 
building could be recorded from 5 stations (Fig. 7) with 8 scans 
in 4 hours. The scanning resolution was in the range of 2 - 3cm. 
The measuring distances to the object were between 24m and 
60m. In total 4 million points were scanned, which yielded a 
file size of 102 MB. 
A^ 
| 
Holstenstraße 
   
Holstentorplatz 
  
Figure 7. Laser scanner stations at the Holstentor in Liibeck 
5.2 Data Processing 
The registration of all point clouds (Fig. 8) was carried out 
exclusively using suitable overlap areas in each point cloud. 
Thereby a precision in the transformation of 8mm could be 
obtained. 
   
and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
The second step of the processing was the generation of 2D 
facade maps. The point cloud could be processed with the plug- 
in CloudWorx using arbitrary cuttings in AutoCAD. For the 
generation of the maps 2D polylines were drawn on the parallel 
projected part of the point cloud. As a result a view of the 
western front facade, a ground plan and a cutting at a map scale 
of 1:250 is illustrated in Fig. 11. A precision of 2em could be 
derived for the 2D processing by comparing known distances to 
the point cloud. Here, the processing time needed for the 
generation of the 2D products took 17 hours in total. 
In a further evaluation a rough 3D block model was generated 
manually from the point cloud in AutoCAD (Fig. 9). For this 
3D model generation the focus was on the representation of the 
main characteristics of the Holstentor. Although deviations of 
up to 10cm to the point cloud were caused by the generalization 
in the modelling process, the 3D model of the Holstentor should 
be recognized in general. Such a reduced and texture mapped 3 
D model could be used for visualization in the Internet or in 
vehicle navigation systems. For the presentation of the data a 
virtual flight can be generated from the point cloud within a 
short time period. The bitmap textures that were generated in 
Cyclone along a virtual camera path were merged to an AVI 
file with a file size of 159MB and a sequence of 13 seconds. 
Due to suitable compression techniques a MPEG video with a 
file size of 3MB could be generated. 
As an example it could be shown using a window (Fig. 10) of 
the Holstentor, that a detail construction from the point cloud is 
generally possible, if a point cloud with a sufficient density is 
available. In comparison to reference distances the accuracy 
was 2em in the 3D CAD model. After additional manual 
operating it is feasible to bind single details to a total model in 
AutoCAD. 
  
Figure 8. 3D point cloud in Cyclone. 
945 
Figure 9. 3D model in AutoCAD ' Figure 10. CAD 3D model of a window.