Abstract
|
|
This project uses a public web camera to track where drivers stop
at a intersection relative to the white stop line. By gathering
data over time, a pattern emerges revealing a norm for drivers.
If drivers are outside of the
norm, this could be used to signal an event close to the intersection.
|
Introduction
|
|
Web cameras are increasingly used for monitoring traffic on freeways
and at major intersections. Many of these cameras are already
installed in major cities to catch red light runners, but we could also
use these cameras to observe local events. An anomaly in the data
could signal authorities to a local event at the intersection, such as
an accident or an event that people are hurrying to/from.
|
Methods
|
|
The method is to count all the cars stopped at an intersection due to
a red light. These cars are classified by where they stopped relative
to the white stop line. The data is gathered in 10 minute increments
several times thoughout many days. We average together data that was
gathered around the same time (within 1.5 hour periods) in order to
examine the validity of the data.
|
Results
|
|
The data indicates that a clear pattern occurs throughout the course
of a day; this can be seen in the topmost figure below. However, we did not have
enough data to show that this pattern was statistically significant.
The bottom figure shows the one, two, and three standard deviations for cars
stopped behind the line. With more data points, we hope that the
standard deviation will improve. Even so, our standard deviation was
small enough to pick out the day and time of CMU's annual Carnival
festival being held a few blocks away.
|
Privacy
|
|
Many "early warning systems" for disasters most compromise some
privacy in order to gather necessary data. This work is significant
because it does not require a privacy compromise to gather significant
data. The cars do not need to be tracked from one camera to another in
order to gather useful data, and we do not need to know anything about
the car or driver. The only information required is the proportion of
cars that stopped behind, on, or past the white stop line in a given
time period.
|
References
|
|
Camera: https://www.discover.pitt.edu/tour/cl_cam.html
(point camera at
intersection of Fifth and Bigelow, lower left portion of camera's
viewing range)
Project Paper (PDF)
|
