filled 
a. The 
ved by 
itional 
n than 
>rview 
üilt-up 
of the 
p, and 
2, new 
of the 
-called 
isizing 
towers 
  
litional 
d with 
nctions 
Arrows, 
on the 
In this 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
  
paper we present the development of the mentioned system. 
The investigation consists of three main parts: 
e Building a test path in a CAD/GIS environment, 
which supports basic GIS functions (for example 
solving the shortest path problem in urban areas) and 
has an application programming interface (APT). 
e The field work (taking photos about road junctions 
with a high resolution commercial camera, and 
positioning by GPS) 
e Application development for continuously presenting 
the images as navigation instructions along the 
calculated path. 
2. DATABASE CREATION 
2.1 Topology building 
At first we got the digital large-scale topographic map. It shows 
the neighborhood of the Budapest University of Technology 
and Economics. We inserted the raster image into the model 
space of the Hungarian reference system (called EOV). We 
used a second-order polynomial transformation to eliminate the 
distortion of the digital image, which caused by the scale 
difference. 
The next step was the digitizing of the road network in order to 
build the digital representation of the network. We used 
directed graphs to describe the test network. Every lane is 
represented as a weighted edge of the graph. More complex 
problem is the modeling of the junctions (Jiang 2002). 
Originally the developers tried to solve the turning problems in 
junctions by expanding the network describing all turning 
directions with edges (Kirby 1969). In this solution the 
junctions are represented as sub-networks consist of more edges 
and nodes as shown in Figure 2. The disadvantages of this 
solution are the increased computational time and computer 
memory need. Others solved the turning direction problem 
without changing the topology of the network. For example 
Ziliaskopoulus and Mahmassani suggest an Extended Forward 
Star Structure (EFSS) to model the junctions. We used the first 
solution because of its simplicity. Figure 3 shows the test 
network in AutoCAD Map. 
% 
Ld 
  
Figure 2. Junction modeling by expanding the network (Kirby 
1969) 
After we had edited the network in the CAD environment, we 
had to build the network topology. Three different types of 
topology are known in AutoCAD Map software: the node, the 
network, and the polygon topology. These topologies are built 
389 
up by three different geometrical primitives: nodes, edges, and 
polygons. In our research we have applied the network 
topology. It is based on the node-edge relations. In this 
particular case the digitized lanes are the edges, and the nodes 
establish the connections between the edges. 
  
  
Figure 3. The test path in Budapest 
The information of the network topology are stored in a 
relational database, which is linked to the certain geometrical 
primitives. The structure of the relational database of edges is 
shown in Table 1. 
  
Inverse 
resistance 
Direct 
resistance 
Start End ; - 
ID Direction 
node node 
  
  
  
  
  
  
  
  
Meanings of the fields are the following: 
e ID: unambiguous identifier 
e Start/End node: pointer of the start/end node of the 
edge 
e Direction: direction of the current edge. If the 
direction of the edge and lane is the same, then the 
value is 1, otherwise it is —1. If the direction does not 
depend on the editing, then the value is 0. The initial 
value is 0. 
e Direct and inverse resistance: the ,,cost” of the edge. 
The shortest path algorithms use this resistance to weight the 
edges. The initial value of the weights is the length of the edges. 
2.2 Field work 
The next step was taking photographs about the junctions from 
the driver’s viewpoint. The images were taken at a resolution of 
2048*1536. The location of the exposure was measured by a 
single geodetic GPS receiver (Leica SR530). Thereafter we 
inserted automatically the photos in the model space as blocks 
by their coordinates. The filename and the path of the files are 
stored as the attribute of the blocks.