Ted Morris, Jory Schwach, and Panos Michalopoulos with the data collector
Engineers needing to measure intersection traffic volumes often have few tools at their disposal. This means many must resort to doing things the old-fashioned way: by counting.
While both temporary and permanent devices to measure traffic data do exist, issues with their intrusiveness and difficult installation are some factors that often limit or prevent their use. Tubular counters such as the JAMAR Trax series, for example, record vehicle counts by sending air pulses through rubber tubes mounted on the roadway. Although the cost of the system is relatively inexpensive, its drawbacks are its intrusive setup, inability to detect turning movements, and susceptibility to damage by large vehicles. Light Detection and Ranging (LIDAR) and radar sensors are nonintrusive, but they detect only a single roadway location, cannot collect turning movement data, and are very expensive.
As a result, manual traffic data measurements done on-site using a push-button apparatus to record each passing vehicle are the most frequent form of data collection used in traffic engineering studies of intersections and arterial streets, especially those measuring turning volumes. Such data collections are not only prone to error but also often prohibitively expensive and time-consuming. And data collected by manual counting or the other sensors mentioned lack a visual, historical record that would allow researchers to go back and investigate or correct traffic measurement errors.
Although it would be beneficial for transportation officials to perform data-collection studies—essential components for tasks such as retiming traffic signals—on a regular basis, the lack of availability of a low-cost, easily deployable, and nonintrusive data-collecting device means that collection is usually completed only when absolutely necessary—such as for an intersection with unusually high crash rates.
Researchers from the University of Minnesota’s Department of Civil Engineering and the Minnesota Traffic Observatory (MTO) have completed a project they hope will change that. Their work, focusing on a low-cost, portable traffic data-collection system, included the design, creation, and deployment of a video-based device that can be used for temporary data gathering and video recording of vehicle movements at intersections and arterial roadways. Data collected from this device can help in minimizing delays and congestion levels while improving overall coordination.
Civil engineering professor Panos Michalopoulos, principal investigator on the project, along with MTO manager Ted Morris and graduate student Jory Schwach, began working in August 2007 to develop and test the device, which can gather data on traffic volume, speed, vehicle classification, turning movements, queue size, conflicting movements, and time headways. Most important, the video system also provides a visual record of traffic characteristics.
The prototype data collector, consisting of a camera mounted on a self-raising mast, is capable of elevating 28 to 30 feet above a road’s surface. The mast, attached to a custom-fabricated base, can be secured to signs, luminaires, or traffic signal poles. Approximately 40 hours of traffic video can be stored before battery recharging or swapping needs to take place, which allows the device to be left unattended for substantial periods of time.
Another of the device’s advantages is that a single unit can cover an intersection of up to five lanes per incoming approach, or a total of 20 incoming lanes. This means that the device can be deployed at almost any intersection. Its small footprint also makes it optimal for urban areas where limited space is a primary concern, Morris says.
To test the device’s accuracy, the team deployed the system at five sites and left it unattended for one to two weeks at each location. The sites included single- and double-lane mixed approaches as well as wide arterials with isolated and mixed turn lanes. Video recordings were collected during morning, mid-day, and evening rush hours in all weather conditions. Algorithms were then designed and implemented on the machine vision system to extract approach counts or speed from the recorded videos in the lab. A subsequent analysis, which compared the processed data from the machine vision system with the ground truth data observed from the collected video, showed that extraction of vehicle counts and mid-block vehicle speeds is feasible. To improve counting accuracy, improvements to machine vision detectors that sense direction of movement of vehicles are required, as well as deploying a second system for the large intersections to reduce effects of across-lane occlusions of overlapping vehicles in adjacent lanes.
Overall, the research team concludes that this system could successfully function as a temporary data-collection tool, offering transportation practitioners, planners, and researchers a functional and affordable alternative to in situ manual measurements and other data-collection devices. The visual record collected by the device—something not available from other traffic data-collection methods—proved to be a valuable tool. The video can be used for additional analysis and research, leading to improved safety and control practices at all types of intersections. For example, Morris notes the device’s usefulness for capturing irregular and infrequent traffic events outside the regular scope of traffic data collection, including pedestrian/vehicle conflicts, crashes and near misses, and gap selection by turning vehicles.
Another possible use for this system, Morris says, is for monitoring traffic conditions in real-time. In 2009, the research team adapted and tested the prototype for detecting rapidly stopped vehicles resulting from vehicles traveling upstream entering work zones, and preliminary results showed it is feasible, Morris says. “The system could also be used in the general transportation security and surveillance field, providing mobile surveillance at trouble spots or special events to first responders and law enforcement,” Morris says.