Abstract:

Intersection traffic operation evaluation parameters, such as queue length, delay, are difficult to be detected directly. A method is presented to synthetically detect intersection operation evaluation parameters based on video processing. First, the queue vehicles are quickly detected by introducing selective background update mechanism for block- level background subtraction combined with block- level frame difference. Then, the first and the last queue vehicles position can be determined by using virtual parallel line, and the number of vehicles queuing can be detected by combining the first and the last queue vehicles tracking with the virtual loop count near the stop line, as a result, other traffic operation evaluation parameters, such as delay, stop times, also can be gained. The test result indicates that the method can effectively detect multiple parameters simultaneously, also can effectively improve detection accuracy rate. Compared with the actual value，the mean error of measured parameters value is less than 5% which can better meet the need of intersection operating performance evaluation.

Key words: intelligent transportation, operation evaluation parameters, video processing, intersection, virtual loop detection, the first and the end queue vehicles tracking

摘要：

针对交叉口排队长度和延误等交通运行评价参数很难直接被检测到的问 题，本文提出了一种基于视频处理的交叉口运行评价参数的综合检测方法.该方法首先 将选择性背景更新机制引入块级背景差分，并结合块级帧间差分实现快速的排队车辆 检测.然后，利用虚拟平行线定位每车道排队车辆首尾端，并结合排队首尾车的跟踪与停 车线处虚拟线圈的计数来检测排队车辆数，进而获得延误、停车次数等其它交通运行评 价参数.实际测试结果表明，该方法能有效实现多参数的同时检测，并且能有效提高检测 的准确率.与实际值比较，参数检测值平均误差小于5%，能较好地满足交叉口运行性能 评价需要.

关键词: 智能交通, 运行评价参数, 视频处理, 交叉口, 虚拟线圈检测, 排队首尾车跟踪