This proposal addresses the problem of bicycle counting. As an alternative to mechanical counters, we plan to use a portable time-lapse video system to acquire several video segments from various bicycle routes. We also plan to develop computer vision algorithms which will detect and count bicycles in these video segments. Our approach will be model-based since in this way we can efficiently separate bicycles from the background. Bicycles have certain shapes and our algorithms will use a database of these shapes in order to detect bicycles. Various issues of reliability, robustness, and safety will be addressed.
- Project number: 1997013
- Start date: 08/1997
- Project status: Completed
- Research area: Transportation Safety and Traffic Flow
Bicycling, Data and modeling