L 
reated a 
shooting 
building 
era and 
enna on 
itioning. 
models 
building 
rface for 
| texture 
ates that 
can be 
e image 
imaging 
m. In an 
along a 
about 20 
about 2 
 10-fold 
  
tion that 
his ideal 
) uneven 
will also 
space in 
apes and 
system to 
throughs 
   
    
  
  
Uneven 
fetails 
Mapped 
texture 
" Shape of buildings 
Figure 9. Ideal 3D model of building 
5. SUMMARY 
We have developed a mobile mapping system that can 
synchronously acquire images and positional information using 
position-sensing equipment and two omnidirectional cameras 
spaced 5 meters apart. This system has the following features. 
An urban environment will have areas in which positioning by 
GPS will not be possible due to shielding effects caused by tall 
buildings. This system has therefore been designed to acquire 
positional information with apparently high GPS accuracy even 
in places where GPS positioning cannot be performed. This 
positional information is synchronized with omnidirectional 
images and enables reconstructed building models to be 
accurately converted to global coordinates. 
A particular advantage of this system is that once offset 
between GPS system position and a camera has been acquired 
from an image, it can be applied to all images. In addition, 
camera attitude information at the time of shooting can be easily 
obtained since this information can be approximately 
substituted by the FOE of the omnidirectional image in question. 
A shutter control function enables omnidirectional images to 
be acquired for any baseline distance and for models to be 
obtained by selecting an optimal baseline distance according to 
height of the target building and its distance from the camera. 
The omnidirectional cameras of the system can capture high- 
resolution omnidirectional images enabling high-quality texture 
to be obtained for mapping to building models. 
Constructing a mobile mapping system that can synchronously 
acquire omnidirectional images and positional information in 
the above way has improved mobility in terms of model 
acquisition and has enabled image capturing at a rate of four 
square kilometers per hour. These achievements mean that the 
time from data acquisition to model reconstruction can be 
shortened by about 2/3 and that personnel expenses and other 
costs can be held down. 
A future issue is finding ways of making this mobile mapping 
system even more accurate with even higher levels of quality. 
This might be accomplished, for example, by raising the 
resolution of omnidirectional cameras through the use of even 
better high-definition digital cameras, or by introducing a 
position acquisition system of even greater accuracy such as 
Virtual Reference Stations (VRS). With these improvements, 
our eventual aim is to raise the accuracy of acquired models to 
several centimeters. 
As for the future of this system, we can envision the 
installation of cameras at positions lower than building roofs to 
enable the reconstruction of the lower parts of buildings. Also, 
by making the entire system compact and mounting it on 
general automobiles, it might be possible to model a wide area 
in a short time and to also reconstruct urban models 
corresponding to changes over time such as seasonal changes. 
REFERENCE 
Ellum, C. and El-Sheimy, N., 2002. LAND-BASED MOBILE 
MAPPING SYSTEMS, Photogrammetric Engineering & 
Remote Sensing, pp. 13-28. 
GIS Development, 2000. http://Www.gisdevelopment.net/ 
(accessed 10 Jun. 2002) 
Horiguchi, S. et al., 1999, Recovering 3D urban model using 
Range data and Sequential Images, Working-shop on Urban 
Multi-Media/3D Mapping. 
Ishikawa, Y. et al, 2001, Estimate Aerial-Image Parameter using 
SCP, IEICE Symposium. (In Japanese) 
Kawasaki, H. et al., 2001. Super-Resolution of Omni Camera 
Image Using Spatio-Temporal Analysis, IEICE Trans., Vol. 
J84-D-II, No.8, pp.1891-1902 
Miyagawa, I. et al., 2000. Shape Recovery from Aerial Image 
using  Factorization Method with Sensor Information. 
ACCV2000. 
Okutomi, M. and Kanade, T., 1993. A Multiple-baseline Stereo, 
IEEE Trans. Pattern Analysis and Machine Intelligence, Vol.15, 
No.4, pp.353-363 
Tamura, T. et al, 1998. The GPS/INS Integration and 
Kinematic Photogrammetry for Mobile Mapping System, ISPRS 
Commission V Symposium. 
Teller, S. et al, 1999. MIT City Scanning Project: Fully 
Automated Model Acquisition in Urban Areas, 
http://city.lcs.mit.edu/city.htm] (accessed 9 Jun. 2002) 
Uehara, M. and Zen, H., 2000. From Digital Map to 3D Map: 
Creating 3D Map by Motion Stereo Utilizing 2D Map, IAPR 
Workshop on Machine Vision Applications, pp.592-595 
Virtual Helsinki, 1999. http://www.virtualhelsinki.net/ 
(accessed 9 Jun. 2002) 
Zhao, H. and Shibasaki, R., 2000. High Accurate Positioning 
and Mapping in Urban Area using Laser Range Scanner, 
Working-shop on Urban Multi-Media/3D Mapping. 
—203—