Geoscience Reference
In-Depth Information
The data set presented here is relatively simple and results are still limited
as only selective omission is considered. Line deformation methods for
smoothing, aggregation, displacement and enlargement are still yet to be
developed according to nautical chart constraints so as to be evaluated.
5- Conclusion and perspectives
This article dealt with the problem of isobath generalization for nautical
charts. Due to the safety constraint specific to this kind of maps, traditional
contour line generalization methods cannot be used. Lines have to be gen-
eralized based on the type of feature they belong to in order to emphasize
potential risks for navigation. As a consequence, different types of opera-
tions have to be applied and combined together. This paper introduced a
novel approach where generalization is feature-driven and operations on
features and lines are coordinated by an agent model. First, the definition
of terrain features and their representation in a feature tree was presented.
Feature tree classifies terrain into peak or pit. The feature tree not only
presents the structure of terrain, but also maintains the shape of slope dur-
ing generalization transformation. Second, the agent model was described.
Two kinds of agents are considered: feature agents at macro-level and line
agents at micro-level. Feature agents hold the information about the mor-
phology (such as the type of feature and its size) and pass them to lines so
that different actions can be performed and evaluated. Constraints consid-
ered for evaluation are the safety constraint, the legibility constraint and
the shape preservation constraint. Based on these, each operation can be
given a score so as to choose the most efficient.
Preliminary results are presented however further work is still required in
this direction as specific methods for line deformation are still to be
implemented. Line smoothing and displacement methods can be based on
(Guilbert and Saux 2008). Enlargement and aggregation operations should
also be introduced. The method should be tested on several maps at differ-
ent scales in order to assess different tolerance values and adjust the
evaluation method.
Two directions are identified for future work. First, feature definition is
based on the analysis of a contour tree. However, the contour tree relies on
inclusion relationships between contours and inconsistencies can appear
when dealing with open contours on complex terrains. One objective is
