International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
Inter: 
  
    
g 
Arable land 
3 Forest 
Heath, 
Me ada: 
Sand 
Other land 
Buffer po*ygor 
A i 
Road segment 
Figure 2 The buffer polygon for landscape attraction 
calculation 
ROAD CLUSTERING 
Besides some attraction points (Points of Interests — 
POIs), a tourist route should also pass through as 
much as possible those road segments with high 
attraction values. Therefore, we first select those 
road segments with high values, for example, > 75 
and then group them based on their geographic 
closeness. 
This process is called c/ustering and those features 
that are close together form one cluster. Thus, a 
cluster includes road segments that have high 
attraction values and that are also close in 
geographic space. The dark lines in Figure 3 shows 
an example of the result of this clustering process, 
in which 22 clusters are created. 
    
Legend E 
(not clustered 
len center : 
—7 
—10 
> —n 
— 17 
E: 
Q Boe = «Er 
x € - 
= $ 
s e 
+ 
S 
ef 
X 
~ 
A 
ej Y. V # 
> 
\ ® 
* à 
Figure 3 Road segment clustering 
After clustering, a point is created for each cluster 
that represents the location, the weight and the cost 
values of the cluster. This is achieved by, using the 
GIS spatial analysis operations, generating the 
centroid of the bounding polygons of each clustered 
feature. These centroids are shown as round dots in 
Figure 3. 
These cluster centres together with the existing POI 
features that have weight or attraction values greater 
or equal to a predefined value (50, for example) | 
become the candidate sites, call it “candidate list”. 
The elements in the candidate list will participate in 
the selection process to construct tourist routes in 
the next sections. 
PATH FINDINGS 
One final step in the data preparation is to generate 
paths between any pair of nodes in the candidate 
list. The output of this process is a new complete 
graph, in which nodes represent the candidate 
tourist sites and edges represent the paths between 
pairs of tourist sites. 
To reduce the complexity, the shortest path finding 
function in the GIS software is used to generate and 
record a path between any pair of nodes in the 
candidate list. However, the objective of the tour 
route algorithm is to maximize the attraction value 
while maintaining the distance, time or other cost 
constraints. This objective is taken into 
consideration during the shortest path generation by 
using the weight attribute that represents both the 
attraction value and the cost value. It is calculated 
by 
Cost 
totalAttractionValue 
  
roadSegmentWeight = 
where rotalAttractionValue is calculated in the 
previous steps, and Cost can be distance, time or 
other cost factors. With this weight attribute, the 
shortest path function is more likely to select the 
road segments that have higher attraction and lower 
cost values. Figure 4 shows an example of such a | 
complete graph, where the nodes represent the 
attractions points and the edges represent the path 
between a pair of them. 
While generating the paths, the program also 
calculates the total attraction and the total cost 
values for each path and these data are stored in the 
database to be used during the route construction 
phase described next. 
2 
+ 
Legend 
bn Ep] Tourist sites 
kj V] paths between tourist sites 
  
- 2-3 
a 2-4 ) 
2-5 
3-4 
— 3-5 
— 4-5 
= road segments ) 
) 
) 
= ; 
Figure 4 Path generation between pairs of 
candidate sites )