Research Improves Traffic Monitoring

Since 2014, there has been a steady flow of about 250 million cars and trucks on U.S roads. That number is expected to rise as the nation’s population continues to grow. As a result, drivers may expect an increase in traffic and commute times. State and local agencies understand the importance of using traffic sensing technology to collect traffic data to help them better manage traffic congestion, road construction, and traffic signal intervals. 

For this reason, the Illinois Department of Transportation (IDOT) teamed up with the Illinois Center for Transportation (ICT), and researchers from the University of Illinois at Chicago (UIC), to conduct two research studies that examined effective ways to both collect and produce traffic data using several video sources. Both the state of Illinois and the city of Chicago have a large number of traffic cameras and other similar devices. However, those devices do not automatically produce accurate traffic and turn counts. Projects  R27-131 “Leveraging Traffic and Surveillance Video Cameras for Urban Traffic” and R27-169, “Opportunistic Traffic Sensing Using Existing Video Sources (Phase II)” sought to address this issue by developing new research results and software to help get accurate traffic counts.

Dr. Jacob Eriksson served as principle investigator on these projects. Eriksson is an associate professor in the Department of Computer Science at the University of Illinois at Chicago. “The goal of these research projects was to develop effective techniques and tools for automatically tracking and counting vehicles that appear in surveillance video. Compared to prior work on video sensing, we explicitly focused on existing video sources, rather than special-purpose cameras. This means we have to deal with a much greater variation, particularly in perspective and video quality. We have developed several new algorithms for vehicle tracking in such `opportunistic surveillance video’, and developed a vehicle counting web portal for traffic analysis” Erikson said. Although much has been accomplished, Eriksson believes that accurate and robust algorithms need to be  developed in the near future.

A Clustering Method that includes ridges, a centerline, and an entry/ exit point.

The goal of Phase I, “Leveraging Traffic and Surveillance Video Cameras for Urban Traffic”  was to investigate the use of traffic and other surveillance cameras for the long-term real-time collection of traffic statistics. A prototype was also developed which demonstrated the ability to perform those tasks under good conditions. The research team also developed a website featuring an explanation of the project and videos showing how the software works.

The Phase II research project, “Opportunistic Traffic Sensing Using Existing Video Sources (Phase II)”, lasted three years and was documented in two reports. The first was an interim report, which discussed how automatic traffic sensing works and how it could potentially yield the ability to produce continuous, daily traffic counts where such video sources exist, as compared to the occasional traffic studies performed today.

The final report documents the adjustments made to the algorithms used in phase one. This would improve accuracy and robustness to adverse environmental and lighting conditions. This phase also introduced a web portal interface for using the system remotely.

The project was guided by a Technical Review Panel chaired by William Morgan, Planning & Systems Section Chief at IDOT. According to Morgan, when Phase I of this project first started, the opportunities to extract traffic count data from video were limited and IDOT was not able to utilize the video data from existing traffic cameras at intersections. Over the course of these research projects, professor Eriksson and his team validated that the existing traffic camera videos could be used to extract traffic count data.

“This research project confirmed our original expectations and through both phases of the research project we have seen improvements in the ability to track vehicle movements and develop traffic count data. We have also seen private sector vendors expand on these capabilities and add functionality with their traffic camera systems to incorporate traffic counting functionality,” Morgan added.


Posted September 2018