Geoscience Reference
In-Depth Information
3 GPS Data and Implementations
The GPS data used in this paper were collected in the urban area of Eindhoven, The
Netherlands. One of the main purposes of this GPS data collection is to
nd
valuable merits of peoples activity-travel trajectories in a long term. Every partic-
ipant was required to join the survey for a period of three months. People carried the
GPS logger, downloaded and uploaded their GPS data to the website. The data was
then processed using the program of TraceAnnotator to generate the possible
activity and travel data. Details about the TraceAnnotator can refer to the paper by
Moiseeva et al. [
13
]. A web-based prompted recall procedure, which allows people
to validate their historical data by modi
filling, removing or inserting in
case of missing or incorrect data, was adopted. Additionally, respondents were also
asked to provide some demography information and the locations where they visit
frequently.
Individuals carried the GPS logger, named Bluetooth A++ Pro. The device has a
good sensor embedded, which can receive well signals within trains. This capability
decreases in some extent the noise induced by signals in urban area, however, the
overall accuracy of the GPS data is in general rather acceptable especially for
traveling data, like car for example. The GPS devices were con
cation,
gured to record
data in every 3 s. The recorded information includes: date, time, longitude, latitude,
speed, distance, accuracy of the measurement (like PDOP, HDOP, VDOP etc.), and
number of satellites.
Although the data collected in the Eindhoven region was targeted as a large scale
of research projects where all available transportation modes and various activity
types were included, in this paper, we are especially interested in the trips by car to
test the ef
ciency of the proposed map matching algorithm. In addition, since the
training of the BBN model needs some facts that can determine the conditional
probabilities among the input variables and the road segment, we extract the trips
by car, which have detailed recorded diary.
The road network data includes all the road segments of the whole Netherlands.
There are three categories of roads: the national road, the provincial road and the
local road. Figure
4
shows the distribution of the GPS log points and road cate-
gories. As a consequence, a total of 10 trips, including 1,227 GPS points are
selected as a test sample.
In order to evaluate the ef
ciency of the proposed method, we labelled in the
dataset the true road segments for every GPS log point using GIS. A
filter was
designed speci
cally to search the set of candidate road segments for a GPS
coordinates. More speci
cally, a buffer with the given radius (di) for a GPS point i
was created based on the coordinates, and the road segments which touch with the
buffer were taken into the choice set.
To determine a feasible value of search radius will help improving the searching
ef
uenced by various
indicators like number of satellites, weather, etc., the spatial pattern of the log points
vary even for a same road segments. The empirical values of 2.5 and 11.4 (unit: meter)
ciency of the whole algorithms. Since the GPS traces are in
fl
Search WWH ::
Custom Search