Shoichi Horiguchi 
ed an set extracted from DEM. Gray points are Boundary type entities. The right image shows the corresponding digital 2D 
lough map. It is obvious that characteristic structures are well matched. 
roads, 
1seful 
  
  
a ms A Meet # 
¥ Ov 
Qt 
"m 
— yr x. 
(b) Digital 2D map 
(a) Image of Boundary point set extracted from DEM 
Figure 2. Result of matching DEM with map 
2.2 Extracting Intersection surface model and Reconstructing 
Figure 3 shows how to extract an intersection 
with 4 roads. Top image shows the road 
network on the DEM. A circle means a node, 
that is to say, an intersection. A line means a 
link, i.e. a road. Pl and P2 are lines used to 
acquire the cross section of an intersection. 
The middle image shows typical cross sections 
as extracted by Pl and P2. We use heuristic 
construction knowledge that intersections are 
almost flat and lie between tall buildings. 
Based on this knowledge we acquire the basic 
parameters, that is to say, the radius and 
elevation of each intersection. Intersection 
radius and elevation, Rave and Eave in Figure 
3, are acquired by analyzing the DEM. Radius 
  
(a) Orthographic 
  
  
  
  
  
  
  
  
  
Rave is the average of R1 and R2. The Pl R 
intersection is taken as the polygon formed by > Rove 
r new the 8 points shown. Elevation Eave is the 
average of El and E2. Most intersections in 
urban areas are flat, so our intersection model 3 E1 
is reconstructed as a flat plane. » 
2.3 Extracting Road surface model 2R; 
2 
Figure 4 shows how to extract a road between 
road, two intersections: C and D. Top image shows 
nmon the road network on the DEM. Road length L 
y grid is acquired as the distance between i E2 
efore. intersection C and intersection D. P1 and P2 > 
are the lines used to acquire road cross "m 
ndary sections. We use heuristic construction e 
type knowledge that roads lie between tall buildings (b) Cross section (c) Intersection model 
roach and that roads is virtually flat across their 
ament width. Based on this knowledge, road width, Figure 3. Reconstructing Intersection surface model 
W1 and W2, and road elevation, El and E2, 
point are acquired by analyzing the DEM. Road 
width Wave is the average of Wi (i=1,2,..., n). Road elevation is the data set of Ei (i=1,2,..., n). In this case, data set Ei 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 415