Database Reference
In-Depth Information
random errors are small errors up to
15 meters caused by the satellite orbit,
atmospheric and ionospheric effects, and receiver issues. We should note here
that errors are related to the spatial positions and not to the temporal aspect of
mobility as it is considered highly precise.
In order to identify systematic errors, researchers may resort to visual inspec-
tion in case of small data sets. For that reason, we could use a filtering method
that filters noisy positions by taking advantage of the maximum allowed speed
of a moving object. This threshold/parameter is used in order to determine
whether a reported position from the GPS stream must be considered as noise
and consequently discarded, or kept as a normal record.
On the other hand, random errors are small distortions from the true values.
Their influence is reduced by smoothing methods. In the literature, different
approaches can be found based on Gaussian kernels, where a smoothed spatial
position is the weighted local regression based on past and future positions within
a sliding time window considering the weight as a Gaussian kernel function,
and Kalman filter, which uses measurements observed over time (the positions
coming in the GPS receiver) and predicts positions that tend to be closer to the
true values of the measurements.
±
2.3.2 Map Matching
The previous trajectory cleaning methods are designed for objects moving with-
out any constraint in their movement. However, real-world applications usually
consider objects that are restricted to move within a given spatial network
that is represented as a graph (e.g., road/railway network) (you can find more
information about this topic on Chapter
3
). Other applications may consider
spatio-temporal constraints (e.g., a pedestrian cannot walk at a speed above a
certain limit, usually bats don't fly during the daytime).
For network-constrained trajectories, themap-matching approach refers to the
mapping of a trajectory to the edges and nodes of the network. More precisely,
the general idea is the replacement of each position of the original trajectory by
the point on the network that is the most likely position of the moving object.
From a computational point of view, map-matching methods can be categorized
to online (processing streams of new positions in real time) or offline (when all
positions are available), while both groups can be further classified as
geometric
,
topological
,or
hybrid
methods.
Geometric methods take into consideration the underlying road network and
various distance measures to determine the actual traveled roads. These distance
measurements can be point-to-point (e.g., Euclidian distance), point-to-curve
(e.g., perpendicular distance), or curve-to-curve (e.g., Frechet distance). For
instance, Dijkstra's shortest path algorithm can be used to determine the distance
