Traffic Video Analytics – Case Study Report
- Ganesh Ananthanarayanan ,
- Victor Bahl ,
- Yuanchao Shu ,
- Franz Loewenherz ,
- Daniel Lai ,
- Darcy Akers ,
- Peiwei Cao ,
- Fan Xia ,
- Jiangbo Zhang ,
- Ashley Song
MSR-TR-1970-3 |
Published by Microsoft and City of Bellevue
Traffic safety and traffic congestion represent one of the biggest problems in modern cities, and
improving these is a major promise of smart cities of the future. Microsoft collaborated with the
City of Bellevue to use video camera feeds for traffic video analytics. The partnership’s vision was
to use the widely deployed traffic camera feeds for traffic analytics on road users as opposed to
traditional mechanical or manual approaches. A primary objective of the partnership was to
evaluate if video analytics can produce accurate outputs on live video feeds to produce
actionable insights to inform Vision Zero strategies. Bellevue is committed to implementing a
Vision Zero Action Plan, with the goal of zero traffic deaths and serious injuries by 2030.
The Microsoft team developed a video analytics platform that analyzed videos to produce
directional counts of traffic users (vehicles, bicycles, etc.). The results were aggregated into a
video analytics dashboard, that was deployed at the City of Bellevue’s Traffic Management
Center from Jul 2017 to Nov 2018, and produced live alerts on abnormal traffic volumes. The
dashboard also helped Bellevue transportation planners understand traffic patterns over long
periods of time. For example, it provided the perspective on vehicle and bicycle patterns, prior
to and after construction of a dedicated bicycle lane on 108th Avenue. Finally, the video analytics
system‘s insights on directional volumes and unusual traffic patterns provided an additional tool
for real-time traffic operations in the Bellevue Traffic Management Center.