Database Reference
In-Depth Information
Figure 11.5 3D spatio-temporal pattern of an itinerary and outlier trajectories.
ninth decile is chosen as left limit of the channel for this
C
m
p
i
. The same
process is computed to define the right limit of
m
p
i
. The left (right) limits are
linked according to the time to define the left (right) limit of the spatial channel.
Figure
11.4
presents the limits of the spatial channel in dark grey. Some positions
are visually outside this channel and can be defined as outliers. In the same way,
the temporal channel is defined. Positions of
C
m
p
i
inside the spatial channel
are split into two subsets, late sided and early sided, according to the difference
between relative timestamps of positions and on the median matched position.
The early and the late limits are computed to define the temporal channel of each
C
m
p
i
. Positions outside the spatial channel are not taken into account because
these parts of trajectories including these positions could be shortcuts or detours.
Spatial and temporal channels at each relative time can be combined to create
the spatio-temporal channel, which is then stored in the knowledge database.
Figure
11.3
d illustrates the spatio-temporal channel of the HGT (Figure
11.3
c)
extracted from zone
A
to
F
of the zone graph (Figure
11.3
b). The spatial and
temporal widths change. For example, for the straight part of the pattern, the
spatial width is bigger than the curved part's width.
The spatio-temporal pattern defines five different zones (usual position zone,
right outlier zone, left outlier zone, late outlier zone, and early outlier zone) for
each relative time. This spatio-temporal pattern (median trajectory plus spatio-
temporal channel) is a 2D
C
t enhancement of the box plot concept. It can be
illustrated in 3D using the Z axis to represent the relative time as shown in
Figure
11.5
. The median trajectory is plotted in black; the usual 3D zones
+
