Geography Reference
In-Depth Information
local conditions were used to evaluate the quality of the GPS data and identify the
noise data and decide which tracking points should be filtered. After the GPS data
had been cleaned up, map matching was conducted to map the GPS trajectories to
the road network and land use on the GIS map. One may use the buffer analysis
based on distance from GPS point to road network proposed by Wolf (
2000
)orthe
method suggested by Chung and Shalaby (
2005
) to combine distance and azimuth
data to identify the links traveled. Most of the GPS tracking data in this study match
well with the GIS map. Not much effort on map matching was needed in this case.
13.5.1.3
Activity and Trip Identification
Since the GPS tracking data are continuous spatial and temporal trajectories, trips
and activities need to be identified from these data. Different methods can be used
to identify activities in the outdoor and indoor environments. The time interval
between two consecutive track points caused by signal loss in an indoor environment
is used to identify indoor activities. The departure time and duration of indoor
activities can be identified in its next record when satellite signals are received again.
The distance between two adjacent points can be used as supplement information
to differentiate an indoor activity from traveling by underground transportation. For
example, one may consider a case with the time interval of more than 10 min and
the distance of more than 2,000 m as underground traveling but not a trip end or
activity in the indoor environment. Trip ends for outdoor activities are distinguished
by calculating the density of tracking points. The point analyst tools in ArcMap were
employed to calculate point density and identify trip ends and outdoor activities. A
previous study suggests that tracking points with densities at least twice that of
the others have a better chance to be trip ends. Similar principles were adopted to
identify outdoor activities in this study.
Altogether 713 activities or trips in 157 person-days have been successfully
detected. Comparing to the 859 activities in 168 person-days reported by the 21
participants, an overall successful rate of 83 % was achieved. This is compatible
to that of other studies such as the 81 % successful rate reported by Stopher
et al. (
2006
). Our investigation into the discrepancies between what derived from
the GPS data and that reported by the respondents revealed that there were 14
activities detected from GPS data but not reported by respondents. These were most
likely the short-duration activities under reported by respondents including buying
something at convenient shops; posting a letter; etc. These activities could easily
ignored or forgotten by respondents. This is the so-called under-reporting problem
of the traditional questionnaire-based data collection methods. Nevertheless, some
of those activities might be mistakenly identified from the prolonged waiting time
at public transport station or in a traffic jam. On the other hand, there were in total
160 activities that were reported by respondents but not identified from the GPS
data. This is largely resulted from the signal loss due to the short period of exposure
to outdoor environment. The 11 person-days missed in the GPS data were caused
by the fact that the respondents were not much exposed to the outdoor environment
Search WWH ::
Custom Search