Geoscience Reference
In-Depth Information
lines has been studied for long and several approaches have been devel-
oped. Different criteria are used such as the distance between the lines
(
GÙNgÙ] 2005 /i aQd Sui 2000
), the slope gradient (
0acNaQess aQd SteveQ
2006
) or the medial axis transform (
0atuN et aO 2006
). The objective of
these methods is mostly to improve the legibility and aesthetic of the map
by selecting or simplifying the contours. Local terrain characteristics can
be taken into account but the character of terrain features modeled by con-
tours is not considered and these methods are not applicable for isobath
generalization on a nautical chart. On a chart, features representing a risk
for navigation such as reefs and shoals must be highlighted while channels
representing main routes of navigation should be indicated.
Generalization must be done by considering the types of features modeled
by contours in order to consider navigation constraints. Limited work has
been done on nautical chart generalization. Papers tackling the overall
generalization strategy of nautical charts are (Tsoulos and Stefanakis 1999)
and (Sui et al. 2005). In both cases, generalization is performed as a
sequence of operations and follows condition-action modeling requiring an
exhaustive description of all possible situations. The final result may
strongly depend on the order in which the lines are processed and the
choice of operations that is made (Harrie and Weibel 2007). Another ap-
proach would be to compute an intermediate DTM from the contours and
to generalize this DTM to get the generalized representation of the relief.
However, the reasoning should be done instead on contours as the user
would identify the features based on the sets of contours that appear on the
map.
More recently, other generalization models based on constraints have been
considered. Instead of imposing rules on what should be done, constraints
stress what results should be obtained (Harrie and Weibel 2007). The most
generic strategy is based on multi-agent systems (MAS). The system treats
elements on the map as agents and allows them to decide their own
actions. An important property of multi-agent modeling is that it is generic
as it can integrate any kind of generalization operation and deal with any
kind of constraint.
The main interest for agent modeling in isobath generalization is that
agents can be used to model terrain features so that the process is feature-
driven. Relationships are established between the different features and the
lines so that they can communicate together in order to take into account
constraints expressed at feature level and at line level. Such an approach
