Database Reference
In-Depth Information
Semantic Annotation Platform
road
train
path_way
(walk)
(bus)
(metro)
Points of Interest
(semantic point)
home
office
market
home
road
train
path_way
(walk)
(bus)
(metro)
?
?
?
?
Road Networks
(semantic line)
residential
area
business
area
market
area
residential
area
Landuse Data
(semantic region)
e
1
e
2
e
3
e
4
e
5
e
6
e
7
context episodes
(stop, move)
GPS
Records
Figure 2.6 Annotation for semantic trajectories.
records, we can compute the trajectory episodes (e.g., stops, moves, which
are largely used in the literature to understand the structure of trajectories,
presented in Chapter
1
); afterward, a couple of dedicated annotation algorithms
are provided for enriching trajectories using additional geo-objects and semantic
tags. There are four main technical components for constructing such semantic
trajectories, as follows:
Building trajectory episodes
: The aim is to build trajectory episodes to fur-
ther understand the inner structure of each individual raw trajectory. Trajec-
tory episode is a subsequence of the raw trajectory. Trajectory data points
inside one episode are more or less homogenous (e.g., staying in the same
place, having the same travel speed), though data points in two neighbor-
ing episodes are unrelated. There are different kinds of episodes, such as
Begin, End, Stop, and Move. In addition to these four types of episodes,
additional episodes can be further designed according to the application sce-
narios, for example, specific episodes for representing congestions in traffic.
The core issue here is to design efficient and robust trajectory segmenta-
tion algorithms to find these meaningful episodes. A couple of trajectory
segmentation algorithms are proposed for building trajectory episodes, such
as velocity, density, orientation, and even time-series-based segmentation
methods.
