摘要
基于视频的交通监控系统具有直观明了,安装方便和维护费用相对较低等优点,成为最有前途的大范围采集交通数据的技术,以指导车辆的运行。该文提出一种高效的,实时的通过计算机视觉来进行车辆检测的方法,结合动态背景刷新策略和动态阈值的选择技术,通过比较检测线上象素的灰度值变化来判断车辆是否通过,然后进行时空分析,把一系列检测线图像按照时间序列进行重构,得到全景视觉图,然后进行图像处理,获得具体车辆的参数(宽度,通过检测线时间等)。试验结果显示了该方法的有效性,车辆通过检测线的识别率大于95%,满足了实际的要求。
Video- based traffic monitoring system has several apparent advantages such as easily intervened and lower costs. It is one of the most promising new technologies for large - scale data collection and implementation of vehicle guidance. This paper presents an effective and real - time approach to detect vehicle based on video image. With a dynamic background updating and dynamic threshold selection technique, we detect the presence of vehicles by comparing the gray value of the sample points of vehicle with that of the road. With 2D spatio - temporal analysis, we can get the panoramic view image by combining a sequence of image on the detection llne along time axis. After we use image processing on the PVI, we can get the vehicle＇s parameter （ width, the duration time）. Experiment demonstrates that the algorithm is effective and the vehicle detection rate is larger than 95%,
出处
《计算机仿真》
CSCD
2005年第9期205-207,210,共4页
Computer Simulation
关键词
车辆检测
时空分析
图像处理
Vehicle detection
Spatio- temporal analysis
Image processing
