Geography Reference
In-Depth Information
spatio-temporal clusters, and discuss methods for creation and analysis of these
clusters.
2 Related Work
2.1 Spatio-Temporal Clustering
Many methods were proposed to cluster spatio-temporal data. Trajectory patterns
of moving objects were mined in Giannotti et al. (
2007
) by finding regions-of-
interest where many trajectories intersect with similar travel times. To find these
places, the geographical space was divided into grids and the density of cells was
computed. Then, a sequence mining algorithm was applied on these regions.
Palma et al. (
2008
) proposed a clustering approach based on the DBSCAN
(Ester et al.
1996
) algorithm to find important places in trajectories. The original
concept of point neighborhood used in DBSCAN was changed to allow finding
important places in a single trajectory. According to a new definition, the
important places are places where the speed of an object is considerably slower
than in other parts of the trajectory.
Zheng et al. (
2009a
) proposed a model to infer a user's travel experience and
the interest of a location. In the first step, trajectories of people were divided into
stops and moves. In the second step, density based clustering was applied on stops
using different scales (neighborhood, city, country), by forming a tree-based
hierarchical graph. For every level of the graph, the interestingness of the location
could then be calculated.
Spatial generalization and aggregation of trajectories was proposed in
Andrienko and Andrienko (
2011
). The characteristic points (stops) of trajectories
were discovered. Then, the points were grouped into clusters. The centroids of the
cluster were used for building a Voronoi tessellation (Okabe et al.
2000
). The
resulting Voronoi cells were used as splitting regions of the trajectory.
2.2 Place Semantics
2.2.1 Semantic Enrichment of Movement Data
Alvares et al. (
2007a
,
2007b
) proposed a method of semantic enrichment of
trajectories using the stop-and-move model. The method combines external
geographical features and finds intersections between important places. Ontology-
based semantic enrichment was proposed in Baglioni et al. (
2009
) to interpret
moving patterns.
