Database Reference
In-Depth Information
sampling frequency, even if the trajectory of the ship crosses the zone). There-
fore, it is difficult to identify whether a ship went through a narrow passage,
entered a restricted area, or computed exact minimal distances to the coast (this
requires interpolation and additional computing costs).
Trajectory features are required to query more correctly and efficiently the
AIS database. Further, it allows for distance computation based on polylines
instead of raw positions, route definitions, trajectory comparisons, and clear
identification of passage through an area or a line. Due to the computing limit,
the number of positions for each trajectory must be reduced using a filtering
algorithm in order to apply spatial operators and functions to efficiently answer
end users' questions. This trajectories production stage is located between Steps
2 and 3 of the data-mining and qualification process (Figure
11.2
).
Many approaches can be considered to define a maritime trajectory and
build such trajectories from a sequence of AIS positions. Let's consider the
time-ordered sequence of all AIS positions of a given ship defined by
S
=
{
p
0
,...,p
n
}
. A trajectory
T
of this ship can be defined as a subsequence of
S
so that
T
⊂
S
∧
T
=
(
p
b
,...,p
j
,...,p
e
) where
p
b
stands for the beginning
position of the trajectory and
p
e
for the ending one.
The main matter consists in selecting the beginning and ending positions from
S
in order to create a set of trajectories. These particular positions (considered
as stops) can be identified by the mobile object cinematic (e.g., null speed),
its spatial position (inside a zone of interest), or the position report sampling
rate (transmission gaps). As the position reports from the AIS itself are not
regular and depend on the ship's behavior (Table
11.1
), a simple time and spatial
threshold might not be sufficient to properly detect gaps defining the beginning
and ending positions and split sequences of raw positions into trajectories. So,
dynamic spatial (
δs
) and temporal (
δt
) thresholds should be derived from the
enriched information provided by the AIS, which contains heading
H
p
, speed
S
p
, acceleration
A
p
,andrateofturn
R
p
indications. Such an approach can rely
on the number of missed frame(s) allowed (
n
mf
) and the reporting intervals
expected by the AIS device onboard (Table
11.1
) to define the time (
δt
)and
spatial (
δs
) thresholds. The next position of a trajectory should be transmitted
within
δt
and should be located within a maximum
δs
distance. Otherwise, the
last position is considered as a stop and future positions of the sequence
S
will
be associated with a new trajectory.
Another way to define these stops within a sequence of positions is to rely
on zones of interest, which can be identified in cartographic information or
manually defined by an expert (see Section
11.2.1
). This inevitably changes the
semantic of the trajectory with respect to the previous method. However, such an
approach is suited better to the analysis of maritime mobilities as ships always
have a small number of well-defined origins and destinations (harbor, mooring,
or waiting area). For a more automatic process, such areas can also be created
