International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
can be observed while in figure 9 some cars can be identified 
parked on the street. These images show the oriented raw 
laser data represented in a mapping reference frame. In order 
to obtain useful information these data have to be filtered and 
edited e.g. for extracting and modeling the buildings 
represented in the images. 
          
     
HALE Reh Ur à Hs 
Figure 8: RGB laser points collected in a kinematic survey 
(laser looking to the right side of the van) 
       
       
ME E. A S 
s 5 Bt IP a. "13 ; pd AT Sg A 
Figure 9: RGB laser points collected in a kinematic survey 
(laser looking to the right side of the van) 
The laser can also be mounted horizontally. In this 
configuration the rotation mirror performs horizontal scans 
perpendicular to the direction of the van. This configuration 
has two main applications: a) while the scanner is looking 
down it can be used to model the road surface and b) while 
the scanner is looking up it can be used to model aerial 
infrastructures. Figure 10 shows an example of an horizontal 
scanner. In that case the scanner was looking up while the 
van was mounted on a train and the train overhead power 
cable was surveyed. 
4.1 Comparison with 1:1000 city map 
An urban survey was carried out in kinematic mode with the 
laser vertically mounted (as shown in figure 5) and 
performing vertical scans of the buildings façades. After 
orienting the laser points using the method described in this 
paper, the point clouds were plotted together with the 
available city map that have an accuracy of 20 cm (1.64 c) 
per component. 
994 
  
  
  
  
   
Figure 10: Intensity image from a train overhead power cable 
(laser looking up) 
It is well known that the determination of the trajectory of a 
van inside a city is very difficult due to the constant GPS 
outages. Therefore, the comparison was done on those parts 
of the trajectory where an acceptable orientation was 
computed. From a laser point cloud plotted together with the 
1:1000 line map ten well defined check points were 
measured. Figure 11 illustrates the selection of one well 
defined point at a building roof corner. The coordinates of 
the laser points were compared to the point map coordinates. 
Table 2 shows the comparison of the ten check points 
identified in the 1:1000 line map and in the laser point cloud. 
  
Point-ID dE dN dH [m] 
1 0.07 -0.11 -0.03 
2 -0.24 -0.16 -0.10 
3 0.30 -0.75 -0.15 
4 0.26 -0.41 0.02 
5 -0.07 -0.08 -0.22 
6 05,102 -0.24 -0.02 
7 -0.04 -0.38 -0.15 
8 0.02 -0.31 -0.06 
9 -0.27 -0.28 -0.24 
10 0.16 -0.25 -0.07 
MIN. MEAN MAX. RANGE RMS. 90 [m] 
dE. -0.27--0.02- 0.30. 0.57 0.19...0.18 
dN.-0.75.-0.30 «0.09. 0. 67 0.35. 40.198 
dH -0.24 -0.10 0:02 0.26 0.13:,0.08 
Table 2: Differences of coordinates at the 10 check points 
(laser coordinates — map coordinates) 
As can be observed the results are on the level of 0.18 m in 
Easting, 0.35 m in Northing and 0.13 m in the vertical 
component. As the path studied is not very long the 
systematic difference observed in the North direction is 
assumed to be caused by a remaining error in the trajectory 
determination. 
In 
Fig 
in 
(be