at once by a small number of cameras. The obtained EEPI (see 
Figure 4) can be translated to EPI with simple geometric 
formulas. 
v image-plane for Et 
     
   
view-point / 
^7 image-plane for EEPI 
Z(moving direction 
of viewer) 
Figure 2: Flow of stationary object under straight moving 
depth-map 
   
  
Figure 3: EEPI 
2.2 Construct of data 
A slit image is obtained from the image sequence. The slit 
spatiotemporal plane image integrated vertical lines of both 
sides of each frame of image sequence toward the temporal axis 
(see Figure 4). 
  
Figure 4: Slit image 
Let this slit image corresponds to distance data obtained the 
EEPI analysis above, we call this depth map urban scene. 
3. UTILIZE DIGITAL MAP DATA AND AERIAL 
PHOTOGRAPH ANALYSIS RESULT 
It is difficult to measure 3D data of buildings because of the 
existence of obstacle, such as a tick growth of trees and 
guardrails in the real city environment. Moreover, the results of 
EPI analysis show that on the edge parts of texture alteration the 
3D data of buildings are densely measured meanwhile on the 
parts of less or no alteration 3D data are roughly measured. 
For the reasons given above, it has been difficult to construct 
the city model accurately only from measured data by EPI 
analysis. Digital map covers the 3D measured points. In the 
following section, we explain the matching method utilizing 
boundary information between buildings in order to match the 
3D data with digital map. 
In addition, we describe the technique to make solid shape of a 
building by using a digital map and the aerial photo analysis 
result. 
3.1 Detecting boundary of buildings from depth map 
As the white points show in Figure 5, the 3D measured points 
appear on the parts of the vertical edge on the depth map. These 
parts are equivalent to the steep texture alteration, such as 
boundaries between buildings, windows and doors. 
Therefore, when we make the histogram of measured points in 
the direction of the camera path (Z-axis), the peak of this 
histogram can be likely judged the prospective boundary of 
buildings (see Figure 6). At this time, we utilize the histogram 
of the measured points in the direction of the X and Z-axis with 
the measured points of buildings facing the street. 
    
Figure 5: Measured points by EPI analysis 
The point that 3-dimensional 
information was measured 
  
° . e e 
ol 2 [ ] Q 2 v 
® ° 
et? A a ec * . M 
  
  
  
  
  
HT 
The building boundary pattern from a histogram 
Figure 6: Histogram of measured points 
On the other hand, we make the pattern of the building 
boundary from a digital map like Figure7. We trace the track of 
the vehicle by GPS over the digital map. A perpendicular line is 
—568— 
  
draw 
track 
to m. 
The 
poin 
from 
the p 
We : 
EPI: 
32 
  
We ı 
with 
heigl 
Sow 
the n 
The 1 
uppe 
uppe 
conv 
pictu 
left t 
Scree 
How 
So, v 
proje 
mate