Collection of Vehicle Activity Data by Video Detection for Use in Transportation Planning

Abstract

Advanced traffic management systems allow video image detection to supplement and improve data inputs in transportation modeling efforts. Video detection systems use machine vision technology, the interaction of video cameras, and specialty computer hardware and software to measure traffic. Traffic parameters such as hourly flows, density, vehicle speed, level of service, and other parameters derived from measured and default values are automatically computed. However, the accuracy of video image detection systems is dependent upon factors such as the camera height, location, and angle above the roadway. Environmental factors such as rain, sun intensity, and day/night also affect vehicle detection accuracy. Existing transportation models can benefit from video image detection technology and improved travel demand models can be developed from such data, providing video detection is accurate. This paper examines how transportation models can benefit from video data. A commercially available system is used to collect data from freeway segments in Atlanta, Georgia. The detected vehicle counts, classifications, and average speeds are compared to true counts obtained over the same interval. Differences in these traffic parameters are determined as a function of camera location and site conditions that constrain the accuracy of video detection. The analytical results lead to recommendations on use of video detected traffic parameters in model improvement.