Chicago has one of the largest networks of traffic monitoring equipment in the United States, and this network has the potential to set the stage for entirely new ways of tracking motor and pedestrian traffic in urban environments.
To capitalize on the information already available from the network, the Illinois Department of Transportation (IDOT) instituted a research project in cooperation with the Illinois Center for Transportation (ICT), “Leveraging Traffic and Surveillance Video Cameras for Urban Traffic” (R27-131). The project was led by Jakob Eriksson, assistant professor in the Department of Computer Science, University of Illinois at Chicago.
The State of Illinois, its counties, Chicago, and other municipalities count traffic through various sources each year. Using these videos will provide additional sources of traffic count data without an increase in collection time or equipment expenses. This will allow additional traffic counts to be completed and provide better supporting data on traffic patterns.
William Morgan, Data Management Unit Chief at IDOT, who chaired the Technical Review Panel that oversaw the project, notes that this solution could also provide an opportunity for increased counts along certain routes during and after special events to better understand the impact on travel patterns.
The research team developed proof-of-concept software for turning traffic enforcement tools into something that can help cities understand traffic movement and make their traffic control systems more useful and efficient—and they can do so using existing technology rather than buying and installing new equipment.
“From the beginning,” Eriksson states, “our goal has been to use existing video sources for vehicle sensing, adding value to existing installations rather than requiring new cameras.”
According to Morgan, “We have produced a proof-of-concept technology for tracking and counting vehicles using existing video resources.”
Eriksson adds that the team had major breakthroughs in using visual data to track vehicles in favorable conditions. Because of the limitations of video quality, the challenge remains of extending those benefits to less-than-ideal conditions, such as rain and darkness.
“Another significant challenge,” Eriksson notes, “has been to get up to speed with the rather far-removed field of computer vision.” This problem has been a topic of interest in the research field for more than 50 years. What was once envisioned as a simple problem to solve has been elusive. “A vision challenge that seems perfectly obvious to a person can be an incredibly difficult problem for a computer to solve,” he adds.
This research has laid a strong foundation for the work. “Accuracy is within 90% under favorable conditions, with improvements still ongoing,” Erickson states. The next steps are to improve the robustness of the results in favorable conditions and work on improving results in less than favorable conditions.
The challenges may be great for such work but so are the rewards, and the researchers are confident that they will solve these problems in the future. As Eriksson notes, “Before this project, we relied on radio frequency–based sensors: GPS, Wi-Fi, Bluetooth, and such, but these get rather poor coverage. With video, you can potentially ‘see’ every vehicle passing through the field of view, which has tremendous promise.”
As part of this project, the research team also developed a website featuring an explanation of the project and videos showing how the software works.