中文摘要：

提出了基于视频双截面的交叉口延误检测方法．在交叉口适当的位置设置两个虚拟检测器，分别判定并记录车辆进入和离开交叉口的信息．在判断车辆进入交叉口时，对新进入交叉口的车辆提取车辆的斑块并进行匹配，如果连续8次斑块都匹配，则表示车辆被检测到；如果判定车辆进入交叉口，则使用车辆斑块和Mean—shift算法对车辆实时跟踪，直至车辆离开交叉口，通过记录车辆进入和离开的时间可以实时地提取车辆的延误参数．实例验证结果表明，基于双截面视频延误检测方法，实时延误的检测精度均值可以达到90％以上，同时，检测的流量精度均值可以达到95％以上．

英文摘要：

The paper introduces a new delay detection method based on the video double section trigger. Two virtual detectors were set to record the vehicles＇ arrival and departure information. Upon the vehicles＇ arrival of the intersection, the vehicles blobs were extracted to be tested whether the blobs could be matched. If the blobs could be matched for eight times successively, the new vehicle was thus detected. A real time tracking by the vehicle＇s stain and mean- shift algorithm would not stop until it departed, and parameters of delay were obtained. With the proposed method, over 90% of delay and more than 95% of the traffic flow precision can be detected on real time.