Geoscience Reference
In-Depth Information
deed. From our tests we found out that the generalization of ditches can be
accomplished with a simple selection algorithm, but that the results are not
always good.
A simple example is the case of a group with an even number of ditches:
selecting one ditch every two, the resulting generalized ditches will be
equally spaced from each other (if this was true in the original data) but
will not be distributed evenly on the space that was covered by the original
ditches. In general, a selection algorithm offer less possibility to represent
the generalized data as it is constrained to keep the position of the objects
that are not deleted. On the other hand, a typification algorithm that creates
a completely new representation of the generalized objects could discon-
nect them.
As we said, though, the odds of a group of ditches being connected to the
hydrography network are much less than a group of an even number of
ditches (50% of probability in normal distribution) and we evaluated more
important a good pattern representation that its connectivity.
For this reason, and in general the ability of a typification operator to per-
form a better generalization of patterns, our choice was to try to develop a
typification algorithm for the generalization of ditches.
Moreover this gave us the chance to approach a more challenging research
topic.
Figure 2:
from left to right: original data, only one ditch present, only another ditch pre-
sent, a group of ditches present. No matter which ditch is selected and which are deleted,
it is only the presence of a group of ditches that allows to convey the concept of the pat-
tern.
