Geoscience Reference
In-Depth Information
that we developed: our set up follows some of the ideas and concepts
found in literature but also introduces some original ones.
2- Approach and methods
The generalization process that has been implemented can be divided in
two parts: the first handles the generalization of the semantic information
in order to make it fit the target data model, the second operates on the
representation of the data, in order to make it fit the specifications for the
target scale.
The semantic generalization requires to measure the width of the rivers in
order to re-classify them; this process is followed by an harmonization step
that improves the aesthetic quality of the results and corrects some mis-
classification errors; the generalization of the representation operates
mostly on the geometries, gathering information from the geometric data
and using the enriched data to reconstruct the river courses that are then
pruned by length and density.
2.1 Width measurement
The width of a river, although being a concept easy to understand, is an
ambiguous definition when it has to be transformed in a metric for an algo-
rithm. The width of a river, in fact, is not represented as a single value but
it varies along the course of the whole river. Since in our datasets each
river is divided in sections, each of them comprising both an area and an
edge of the graph passing through it, we decided to measure the width of
each section. This task can be accomplished in many ways, more or less
precise (e.g. with a perimeter and area ratio): our choice was to develop a
simple algorithm, mimicking the manual process of river width measuring.
The algorithm to measure the width of each river section area samples the
edge in
n
points equally spaced
d
meters. For each of these points a line
normal to the direction of the edge in the point is drawn and the first cou-
ple of intersection points with the boundary of the area are found; the
width of the section will be the average of the distances between each couple
of intersection points. In our tests, the distance
d
was set to the value of
15 meters.
