Effects of fog, snow, and rain on video detection systems at intersections

Abstract

The flexibility and adaptability of video detection systems (VDS) make them attractive for vehicle detection at signalized intersections. However, the VDS performance could be affected by illumination and weather factors. This paper presents an evaluation of three commercial VDS under six adverse weather conditions including snow, fog, and rain, during daytime and/or nighttime. The systems were installed side-by-side at a signalized intersection, and evaluated at stop bar and advance zones based on four detection errors: false calls, missed calls, stuck-on calls (detections that are not terminated soon after vehicles depart), and dropped calls. Activation and deactivation timestamps were initially screened using a computer algorithm, and finally all errors were manually verified using video images. Results showed significant changes in the VDS performance under the different weather conditions. False activations increased in most scenarios but to a greater extent in snow conditions during both daytime and nighttime (typically to more than 50%, and up to 90%). Missed calls also increased in snow conditions and during dense fog (typically to more than 10%, and up to 50%) mostly at the advance zones. Stuck-on calls increased mostly in nighttime rain (less than 10%), while in general dropped calls were rare. Detailed quantification of the errors in each detection zone, and their potential causes are presented.