Geoscience Reference
In-Depth Information
Selection is concerned with the semantics of the features rather than their
location attributes and as such represents the abstraction of the symbolic
aspects of the map (Edwardes et al. 2005). Selection involves the identifi-
cation of objects to retain or eliminate from the database (Slocum 2005). In
fact, both locational and semantic attributes must be considered for selec-
tion.
The goal of point selection is also to avoid over-plotting and un-
readability. After point selection the size of the displayed points can also
be increased to achieve a better visibility and legibility. In the past research
in generalization of point data via point selection has focused on the
development of automated and assisted techniques (e.g. Weibel and Jones
1998). The algorithms developed for such applications must be operational
for use with large data volumes, but the absolute levels of performance are
not critical (Burghardt et al. 2004).
In the point selection process the main goal is to keep the information
content (entropy) of the original data as good as possible. What kind of
information exists in point data? In the literature there are four types of
information contained in (point-) map features: statistical, metric, thematic
and topological information (Sukhov 1967 and 1970, Neumann 1994,
Bjørke 1996, Li and Huang 2002).
Statistical information of a map considers all the occurrences of the map
features as unique events and all map events are equally probable. It is
simply computed by counting the number of map features. But the spatial
distribution is not considered. The topological information is considered
by the connectivity and adjacency between map features. Thereby different
types of relations between the map features exist.
Table 1
shows the work
conducted by Yan and Weibel (2008). The metric information considers
the variation of the distance between map features and consequently their
densities.
These information types of points and point cluster can be quantified using
different measures (Ahuja 1982; Ahuja and Tuceryan 1989; Yukio 1997,
Langran and Poicker 1986, Flewelling and Egenhofer 1993, Van Kreveld
et al. 1997).
