城市道路交通异常事件自动检测方法  

Automatic Detection Method of Abnormal Events in City Road Traffic

作　　者：万福才[1] 尹承祥 韩晓微[1] 王东政[2] 

机构地区：[1]沈阳大学信息工程学院,辽宁沈阳110044 [2]大连理工大学电子信息与电气工程学部,辽宁大连116024

出　　处：《沈阳大学学报（自然科学版）》2015年第1期30-33,43,共5页Journal of Shenyang University：Natural Science

基　　金：辽宁省教育厅一般项目(L2012429 2012-2015);沈阳市科技计划项目(Grant No.F12-169-9-00)

摘　　要：针对城市道路交通流量大,易发生交通违法异常行为,提出了一种城市道路交通异常事件自动检测方法.该方法在进行异常事件检测时,首先对道路交通进行视频序列采集,然后提取视频中的车辆加速度、方向变化、几何位置信息.在提取这些特征值后,计算三个特征值变化率指数总和,并与设定的阈值进行比较,判断是否发生异常事件.通过对同一道路的相同监控视频片段采用不同的算法进行检测仿真实验,测试结果表明该方法得到的正确检测率DR为98.6%,误报率FAR为7%,相对于基于临界安全区域方法、基于Boosting检测方法其整体检测性能有所提高.According to the situation that the flow of the city road traffic is large,and the traffic illegal behavior is easy to happen,an automatic detection method for the abnormal events of the city road traffic is presented.When using the method to detect abnormal events,video sequences of the road traffic are collected firstly,and the information of the vehicle acceleration,the direction of change,and geometrical position are extracted secondly.The sum of three feature values index change rate is calculated after extracting the feature values,and compared with the threshold to judge whether abnormal events happen.Different algorithms are used to do the simulation experiment for the video clip of the same road.The detection test results show that the correct detection rate DR of the method proposed is 98.6%,the false positive rate FAR is 7%,and the overall detection performance is improved compared with the critical security region method and the detection method based on Boosting.

关 键 词：异常事件 城市道路交通 加速度 几何位置 自动检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]