Full text: The 3rd ISPRS Workshop on Dynamic and Multi-Dimensional GIS & the 10th Annual Conference of CPGIS on Geoinformatics

```ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS", Bangkok, May 23-25, 2001
167
determined by two factors, i.e. (a) the size of feature and (b) the Voronoi regions to describe spatial relations between map
neighbouring features. Indeed, Chen et al. (2001) have used features.
Figure 3 Voronoi diagrams of the maps shown in Figure 1
For these reasons, the authors try to relate spatial information of
map features to their Voronoi regions to develop a set of new
quantitative measures. However, detailed discussion of the
formation of Voronoi regions is outside the scope of this paper.
Algorithms for generation of Voronoi region in vector mode have
been presented by Okabe, et al. (1992) and a raster-based
algorithm has recently been proposed by Li et al. (1999).
Therefore, no further discussion on this topic will be presented in
this paper.
3.2 Metric information on map: entropy of Voronoi regions
Metric information here considers the space occupied by map
symbols only. In this case, an analogy to the entropy of binary
image is used. That is, if the space occupies by each symbols is
similar, the map has larger amount of information. If the variation
is very large, the amount is smaller. This can be achieved by
using the ratio between Voronoi-region of a map system over the
enclosed area of the whole map as the probability used in the
entropy definition. Let S be the whole area and and it is
tessellated by Si, i=1, 2, ... N. Then, such a probability can be
defined as follows:
The entropy of the metric information, donated as H(M), can then
be defined as follows:
W(M) = H(P,, P 2 P, ) = -£ ^(In S,- - In S) (5)
(=1
H(M) has its maximum when R has the same value for all i=1, 2,
.. .N. In other word, when the area S,is equal. Mathematically,
H(M) m =H(.P„P 2 ,...,P. = (6)
For example, the two maps shown in Figure 4 have different
amount of metric information although both are tessellated by 9
polygons. The map in Figure 4(b) has the maximum H(M) for
any tessellation into 9 polygons.
In the case of map, it is clear that for the same number of
feature, the entropy will be larger is the symbols are more evenly
distributed. However, it is clear that such entropy is related to
the number of map symbols and thus it would be not convenient
to compare two maps with different number of symbols. In order
to overcome this shortcoming, the entropy could be normalised
as follows:
H(M)„ =
h(m)
H{M) m
(?)
Another possible measure is the ratio, Rm, between mean of the
areas (m A ) and the standard deviation a A .
m
l n
,—L4
n
i=l
I
n -1
(8)
(9)
R
M
(10)
Figure 4 Two different tessellation of an area,
resulting in two different amount of metric information
```

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.