Video-Based Advanced Analytics is Detailed Enough to Identify Near-Crashes, Classify Road User Types, and Detect Speeding Infractions and Lane Violations.

Case study from the city of Bellevue, WA details how advanced video monitoring can provide highly detailed data on traffic volumes, road user speeds, and near-crash traffic conflict indicators.

Date Posted

Accelerating Vision Zero with Advanced Video Analytics: Video-based Network-wide Conflict and Speed Analysis

Summary Information

Motivated by the city’s goal of zero traffic incidents in the near future (Vision Zero), the “Video-based Network-wide Conflict and Speed Analysis to Support Vision Zero in Bellevue (WA) United States” project began in August 2019. The city installed a network of 360-degree, high-definition traffic cameras at 40 intersections to collect advanced visual and then employed AI machine-vision algorithms to identify notable events.

In addition to the case study, in July 2020, the city published three reports using the data collected, including one on conflict analysis, one on speeding analysis, and one on the correlation between conflicts, speeding, and crashes.


The city chose to study 40 intersections representing a diverse spread of characteristics including location, land use, density and road geometry. The cameras collected data 16 hours per day for a week in September 2019. Captured was around 5,000 hours of footage, 8.25 million road user observations and 20,000 critical conflict interactions.

Using artificial intelligence algorithms to process traffic camera footage, the project team was able to ascertain traffic volume, road user speed, and near-crash event data. Video-based monitoring has several advantages over traditional methods. For example, video-based monitoring can detect near-crashes, classify road user types (cars, bikes, pedestrians) and their movement paths, as well as detect infractions such as speeding and lane violations. Data collection in the form of video also were easy for human researchers to review and understand, unlike other methods such as LIDAR or Bluetooth sensor data which often provide numerical data.


 Using video-based monitoring and artificial intelligence algorithms for incident detection, the team found many pertinent results. These include:

  • People riding bicycles were 10 times more likely to be involved in a conflict than motorists.
  • More than 10 percent of drivers were speeding, with half of them going more than 11 mph over the posted speed limit.
  • Motorcyclists travel faster and are more likely to be involved with critical conflicts than any other road users.