Database Reference
In-Depth Information
the late area, so this object can be spatially qualified as “inside the channel” but
temporally as “running behind schedule.” Such real-time analysis methods can
be used to predict the destination and time of arrival of the ship once an itinerary
has been matched, and if the position is normal. The destination prediction can
be higher than 90%. In the same way, the confidence interval of time of arrival
could be the width of temporal channel at the arrival.
11.3 Conclusions
The maritime environment represents an increasing potential in terms of mod-
eling, management, and understanding of mobility data. The environment is
typical and recently several real-time positioning systems, such as the Auto-
matic Identification System (AIS), have been developed for keeping track of
vessel movements. This chapter outlines different aspects of maritime mobili-
ties understanding through pattern discovery and analysis of ships' trajectories.
Underlying issues concern in particular trajectory modeling problems, trajectory
querying and simplification, similarity functions, classification and clustering
algorithms, and knowledge discovery (trends, unusual behaviors, and event
detection).
Assuming that moving objects at sea that are following the same itinerary
behave in a similar way (considered as the normality), this chapter illustrates the
different steps leading to outlier detection. The suggestedmethodology considers
several steps. First, the data flowprovided by the automatic identification systems
is managed in structured spatio-temporal databases. Then, data mining processes
are used to extract trajectories (vessels of the same type) and spatio-temporal
patterns between two zones of interest (an origin, a destination). Each pattern
includes a median trajectory and a spatio-temporal channel that describes the
dispersion of the set of trajectories. Such trajectory patterns are meaningful to
understand maritime traffic and detect outlier positions in real time. Indeed,
each new position (partial trajectory) can be spatially and temporally qualified
according to spatial and temporal criteria. For end users monitoring maritime
traffic, such real-time qualification of positions and trajectories is tied with
triggers automatically executed when a new outlier is detected, and adapted
geovisualisation process are essential for safety purposes.
While complete, the suggested methodology still leaves several additional
challenges. First, cartographic information and environmental data such as cur-
rents, tides, and winds that affect ships' movements could be taken into account
for further improvements. Many other algorithmic approaches for trajectory rep-
resentation and reconstruction can be considered for other knowledge discovery
objectives. Interactive and adaptive geovisualisation is also of interest. Another
challenge concerns new itineraries. Many factors can influence ships' behavior,
leading to the apparition of new itineraries. The proposed approach handles such
