Wed May 21, 2014
Big Data & Bike Safety
Urban planners in Virginia are trying to make bicycling safer, but they’re hampered by a lack of statistics about who’s riding where.
Alec Gosse rides his bike to work at a Charlottesville company that analyzes data, and this year he was working on a PhD in environmental engineering. Those interests led him to try and solve a problem daunting city planners.
“There was no data for how many bikes were using various roads in the city. It just didn’t exist.”
Without that information, they didn’t know where to make road improvements for cyclists.
“If you were deciding say between putting a bike lane on one street or on another street – or putting a bike lane on ten streets or building one very fancy off street path.”
Hiring people to count would be very expensive, with so many streets around town, and bike traffic varies from day to day. When UVA is in session, there would probably be more. When snow is falling, there would likey be none. So Gosse came up with a better idea. VDOT and the city have cameras at many major intersections to better synchronize stop lights with traffic. With $40,000 in funding from the University of Virginia Alumni Association, Gosse and another graduate student, Emmanuel Denloye-Ito created software that could review video from those cameras, identify and count bikes. It wasn’t easy.
“They look different from the front and the back and the side, and they’re just not very big.”
But they found a way to do it, and Gosse suspects this software could be refined to make cycling safer. Safety, he reasons, is a function of risk and the number of people exposed.
“So by counting, we could get how many people who are exposed and then you could look for close calls.
That would be the next step. Rather than waiting for someone to actually crash, you record the number of close calls, and you estimate the likelihood that they might crash, and then fix it before they do.”
Last week, he got his degree, but the research continues at UVA, and Gosse hopes there will be government funding to put the software to work.