摘要
文中以上海市城市快速路及其交通监控系统作为研究平台,通过对大量实际事故数据的分析,提出采用非连续滑动1分钟占有率序列作为自动事故检测的基础数据,提出了自动事故检测的五大关键参数和基于动量因子BP神经网络的关键参数阈值预测方法,并最终建立了多参数联合的自动事故检测方法。与基本加州算法、加州7#算法和McMaster算法的比较,表明该方法具有良好的自动事故检测效果。
Based on the advanced traffic management system （ATMS） of Shanghai urban expressway, through analyzing sufficient field incident data, this paper advanced the Un-Continuous Moving One Minute Occupancy Sequence （UCMOMOS） which is used as basic traffic data （BTD） for automatic incident detection （AID）, five AID critical coefficients and the coefficient threshold forecasting method. Ultimately, the incident judgment method of multiple co-operating AID coefficients was put forward. According to actual incident data, based on the parallel test method, this paper evaluated the Multiple Coefficients AID algorithm and several existed AID algorithms. The evaluation result proved the Multiple Coefficients AID algorithm had a good performance.
出处
《物流工程与管理》
2009年第8期147-149,共3页
Logistics Engineering and Management
关键词
城市快速路
自动事故检测
基础数据
关键参数
神经网络
urban expressway
automatic incident detection （AID）
basic data
critical coefficient
