EVALUATION OF UDOT'S VIDEO DETECTION SYSTEMS: SYSTEM'S PERFORMANCE IN VARIOUS TEST CONDITIONS

Abstract

Video detection systems have replaced the industry's standard inductive loops because they are easy to use and install, have flexible applications, and their maintenance is cost effective. These systems involve positioning cameras above traffic to capture images of passing vehicles. These images are analyzed by a vision processor using application specific algorithms. Virtual loops/detection zones are superimposed on these images. The activation of these virtual loops signifies detection. But the performance of this system under different environmental and light conditions is still unknown. The generation of false and missed calls has lowered the accuracy of video detection systems. This study evaluates the performance of video detection systems in Utah. The results also showed that the video detection system performed well under day and dusk conditions with 87.2% correct detection for both conditions. The night condition recorded a correct detection of 73.4%, with 19.9% false calls. Video detection in inclement weather generated 81.3% correct detection and 14.1% false calls. Missed detection under all conditions ranged from 4.6% to 6.8%. Overall, video detection generated 83% correct calls and 17% discrepant calls. Out of the discrepant calls, 14% showed consequences on traffic signal timings. This study recommends paying closer attention to the proper installation of video detection systems. This includes the accurate placement of cameras, sufficient background lighting, focusing settings and field of view calibration. Additionally, detectors need to be placed accurately. Employing vendors for the initial installation of video detection at each intersection, followed by routine checks to ensure that the systems work effectively is also recommended. All vendors' systems should also be tested and compared with inductive loops at a single intersection under different weather conditions. The cost savings associated with using a single vendor may be assessed. This would lead to a higher level of expertise and a lower error rate.