University of Minnesota Driven to Discover
U of MNUniversity of Minnesota
Center for Transportation Studies

Research shows accuracy of video camera-based bicycle counting system

With increasing congestion on roads, alternative methods of transportation are looking more and more popular. Bicycles, for example, offer exercise as well as transportation, and more and more communities are investing in multi-use paths.

Does the investment in such paths result in high use? Communities who are looking for answers to that question may want to consider a new method of counting the number of bicycles on a trail or road.

As a result of research conducted at the University of Minnesota and funded by the Minnesota Department of Transportation, the system can reliably count-up to 70 percent accuracy-the number of bicycles on a trail.

Researchers Nikos Papanikolopoulos and Scott Rogers from the University’s Artificial Intelligence, Robotics, and Vision Laboratory developed a system to monitor bicycle activity in sequences of gray scale images from a video camera. Concepts from the proposed system also could apply to other uses, such as detecting, tracking, and classifying vehicles, pedestrians, rollerbladers, and other traffic objects.

Cameras often provide richer and more complete information than other methods for counting bicycle use, such as loop detectors, laser triggers, and manual counting. Cameras also offer a less intrusive, more mobile method.

This work builds on previous camera-based research, which focused primarily on distinguishing bicycles from the background. This new approach counts each bicycle only once and also makes it possible for detection of bicycles at more difficult angles.

The system uses a simple bicycle model of two circular objects separated by a relatively known distance. This model helps the computer detect the bicycle image.The system identifies raw images, blobs or image regions, edge images, based on the bicycle model. A personal computer processes the raw images.

Researchers successfully tested the system on a dual Pentium computer equipped with a Matrox imaging board. The system achieved a peak performance of eight frames per second. Experimental results based on outdoor scenes show promising results for a variety of weather conditions.

For a copy of the report, Bicycle Counter, call 651/282-2274. For more information about the research project, contact Nikos Papanikolopoulos at 612/625-0163.