摘要: |
Recent years have seen increasing deployment of radar-based technologies for vehicle detection at signalized intersections in the United States, mainly because they are nonintrusive, accurate, and robust to varying lighting, weather, and environmental conditions. In this paper, a radar-based detection technology is evaluated in the context of various weather and environmental conditions. High-resolution (100-ms sampling interval) data were collected in the field from two representative test sites. The detection errors were correlated with varying weather and environmental conditions using data-mining techniques, such as conditional inference trees and regression models. It shown that false and stuck-on call errors tend to increase under more-adverse weather conditions (e.g., rain and thunderstorms). Visibility, glare, and uneven shadows appear to be irrelevant. The near-side mounting location is associated with reduced missed-call, false-call, and dropped-call errors. |