基于视频的交通事件自动检测技术综述

首先从基于视频的交通事件自动检测技术所涉及的目标提取、车辆跟踪和行为理解等3个步骤中，详细地介绍和比较国际上常用的各种检测算法；其次，从背景建模和车辆阴影两个方面分析了检测算法中存在的技术难点和可能的解决方案；同时，从平均检测时间、检测率和误报率3个方面介绍该技术算法性能的评价方法。最后总结了基于视频的交通事件自动检测的相关技术，展望了该技术的发展前景。

Survey of Vision-based Automatic Incident Detection Technology

In recently years,vision-based automatic incident detection(AID) has become a necessity given the ever increasing traffic density for most major intersections and highways.This paper firstly introduces and compares various detection algorithms from three aspects,object detection,vehicle tracking and activity understanding involved in the vision-based AID technology;analyzes the technical difficulties and their possible solutions based on the background modeling and vehicle shadow,and discusses the estimation method of this AID algorithm performance from three aspects,i.e.mean time to detect,detection rate and false alarm rate.Finally,a summary of these techniques that are relative to the vision-based AID is presented,as well as the development frend of the technique.