Database Reference
In-Depth Information
analysis, one may need special techniques that support focusing on the context
items and relationships of interest.
Position records inmovement data may include some context information, but
this is rarely the case. In any case, movement data cannot include all possible
context information. Typically, the source of relevant context information is
one or more additional data sets describing some aspect(s) of the movement
context. We shall shortly call such data “contextual data.” Context data may
result from previous analyses of movement data. In our previous examples we
have demonstrated derivation of spatial events, event clusters, as well as classes
(clusters) of locations and of time moments. Such derived data can be considered
as context data and used in further analysis of movement data.
The general approach is to derive contextual attributes for trajectory positions
by joint processing of movement data and contextual data and then visualize the
attributes to observe patterns and determine relationships. The derived attributes
may characterize the environment (such as weather conditions) at the positions
of the moving objects or relations (such as spatial distance) between the positions
and context items in focus. Values of these attributes are defined, as a rule, for all
trajectory positions. The analyst looks for correlations, dependencies, or, more
generally, stable or frequent correspondences between the contextual attributes
and movement attributes.
Besides stable relationships between movement and its context, the ana-
lyst may also be interested in transitory spatial, temporal, and spatio-temporal
relationships occurring between moving objects and context items during the
movement and lasting for limited time. This includes, in particular, relative
movements of two or more moving objects such as approaching, meeting, pass-
ing, and following, and relative movements with respect to other kinds of spatial
objects. Such occurrent relationships can be regarded as spatial events since they
exist only at certain positions in space and in time.
Many types of relationships can be expressed in terms of spatial and/or
temporal distances. This includes proximity between moving objects, visiting
of certain locations or types of locations, and being in the spatio-temporal
neighborhood of a spatial event. Spatial and/or temporal distances from moving
objects to context items can be computed and attached to trajectory positions
as new attributes, which can be visualized and/or used in further analyses.
Particularly, they can be used for filtering and event extraction as described in
Section
8.3
.
As an example of analyzing movement in context, we shall investigate how
the speed of car movement on motorways is related to the distances between
the cars. Hence, there are two aspects of the movement context in which we are
interested: type of location (specifically, motorway) and other cars (specifically,
distances to them). The distances between the cars can be determined directly
from the trajectory data; no additional data are needed. This can be done using
