Effect of shadows and time of day on performance of video detection systems at signalized intersections

摘要

This paper presents the effect of shadows and time of day on the performance of three video detection systems (VDS): Autoscope, Peek and Iteris, at a signalized intersection. Vendors were given two opportunities to improve the performance of their initial VDS setup before the final data collection. The evaluation and the results are based on 40 h of data from 20 days. Four performance measures: false calls, missed calls, dropped calls, and stuck-on calls were used in this study. Automated analysis of the data was performed in conjunction with manual verification of videos to identify the detection errors. The results for stop bar detection zones are presented in this paper. All the VDS reported false calls and depending on the lane, time of day, and sunny or cloudy conditions, the average false calls varied from 0.2 to 36. The false calls were higher in sunny conditions than in cloudy conditions due to the shadows. Also, false calls for the leftmost lane were higher in sunny morning than in sunny midday due to shadows of turning vehicles from the middle lane. The false calls in the other two lanes were not affected by the time of day. Under the studied conditions, Autoscope, Peek, and Iteris missed detecting 2, 9, and 0 vehicles, respectively, in more than 7000 vehicles that went through the intersection. Autoscope, Peek, and Iteris had 87, 5, and 1 stuck-on calls, respectively, and there were more stuck-on calls in the morning than at midday. None of the three VDS had any dropped calls.