International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
4. EXPERIMENTAL RESULTS AND APPLICATIONS 
IN DATA FUSION 
A VLMS data that is taken in GINZA area, one of the major 
commercial centers in Tokyo, lasted for about 15.7km. Figure 7 
shows vehicle's measurement course using the building facades 
that are represented by the laser points of VLMS. See Figure 12 
for a strip of the line images and a perspective and close view 
of the laser points. A DSM, which was generated from an air- 
borne laser data that has a ground resolution of Im” and a 
ground coverage of about 15.9 km”, is used to rectify the 
GPS/INS parameters at each update. In addition, a 1:2500 
digital map containing the data of 3D building frames only, 
where 2D building frames were generated from aerial photos, 
elevation data for each facade was extracted from an air-borne 
laser data, are exploited to test the performance in data fusion. 
; i 
The building facades measured by M 
the laser points of VLMS UA 
  
Figure 7. The laser points of VLMS at GINZA area and the 
distribution of tie-points for horizontal registration 
10m i J 
bs Tie points 
(a) (b) 
Figure 8. A result of registering the laser points of VLMS with a 
DSM 
4.1 Experimental results of rectifying VLMS data 
Figure 8 shows an overlapping of the three different data 
sources, where large displacements are found from the laser 
points of VLMS to other data sets. In the area "A" of Figure 8, 
vehicle ran along the street a second time after a several ten 
  
408 
minutes. Errors in GPS/INS parameters are accumulated much 
more during this period, so that laser range measurements to the 
same building facades do not match well, and both drift away 
from the data of DSM in different patterns. In horizontal 
registration, 18 sets of tie-points are manually assigned, binding 
the corresponding building corners that are measured by the 
laser points of VLMS and the DSM. Distribution of the 18 sets 
of tie-points is shown in Figure 7. Horizontal parameters at 
more than 46000 GPS/INS updates are corrected within several 
seconds using the 18 sets of tie-points. A result of horizontally 
registering the laser points of VLMS to the DSM is shown in 
Figure 8. 
4.2 Applications of data fusion 
An interface for semi-automatically extracting a broad range of 
urban objects using both laser points and line images was 
proposed, and an application of the interface using VLMS data 
was developed in our previous research (Zhao and Shibasaki 
2003b). Laser points are projected onto line images. Using line 
images as the interface, using laser points for 3D information, 
manually drawing the boundary of the target objects, geometry 
of the objects are automatically calculated from the 
corresponding laser points. Figure 9 shows an example of 
object extraction using the interface. 
On the other hand, building frames of the 1:2500 digital 
map can be projected onto the line images of VLMS by looking 
for the pixel that has the same projection vector with that of 
each building corner. Textures of the building facades are 
generated automatically by projecting and  re-sampling 
corresponding image pixels onto the plane of the building 
facades. A view of the textured buildings, as well as the objects 
that are extracted from the VLMS data is shown in Figure 10. 
5. CONCLUSION 
This paper contributes to a method of fusing the data output of a 
mobile mapping system - VLMS with existing geographic data 
sources, aiming at enriching the database of urban details. An 
algorithm is developed to rectify the GPS/INS parameters that 
might be quite erroneous in urban area by registering the laser 
points of VLMS with an existing data source, e.g. a DSM. The 
algorithm is examined using a VLMS data that are taken at 
GINZA area, Tokyo. The laser points of VLMS are horizontally 
and vertically registered with a DSM, where 18 sets of tie- 
points are manually assigned, and four parameters at each 
GPS/INS update are corrected automatically and efficiently. 
Objects are extracted from the rectified VLMS data, which 
consist of commercial sign board, traffic sign/signal, road 
boundary, road lights etc., and fused with a 1:2500 digital map. 
In addition, line images of VLMS are projected onto the 
building facades of the digital map, and textures are generated 
in an automated way. 
REFERENCE 
[1] Collins, R., A.Hanson and E.Riseman, Site Model 
Acquisition under the UMass RADIUS Project, Proc. of 
Arpa Image Understanding Workshop, pp.351-358, 
November 1994. 
[2] Gruen, A., 1998, TOBAGO - a semi-automated approach 
for the generation of 3-D building models, ISPRS Journal 
of Photogrammetry and Remote Sensing, vol.53, issue 2, 
pp.108-118. 
[3] Li, R, 1997, Mobile mapping: An emerging technology 
for spatial data acquisition, Photogrammetric Engineering 
and Remote Sensing, 63(9), pp.1085-1092. 
Inte 
  
[9] 
[10 
  
Fig 
aut