基于多层前馈网络的高速公路意外事件侦测及规律

为了建立基于高速公路交通流参数检测的事件救援系统,运用多层前馈网络(MLF),对其关键的意外事件侦测技术及其检测规律进行了研究。确定了事件检测需要选定的模型,描述了检测的基本原理,提出了模型的结构原型并进行了训练和测试;对输入特征中的单侧与双侧输入、多时阶组合及多因素组合特征进行了性能测试,得到了相对具有良好性能的输入特征。同时,给出了不同因素及因素组合与检测效率之间的基本规律。结果表明,给出的模型较California 8#算法性能显著改善,其检测率提高了187%,误检率降低了46%,平均检测时间缩短了73%。

China expressway accident detection model and its laws based on multi-layer feed-back network

The accident detection technology was studied by using multi-layer feed-forward model in order to construct an expressway rescue system based on flow parameters.The model for accident detection was designed,the theory of detection was described,and the prototype of detection model was put forward.The input characteristics that include upstream information and downstream information,multi-time interval feature,and composed factors feature,were analyzed.As a result,a good input characteristic was obtained and the law that associates different factors and different composed factors with detection performance was brought forward.The results show that,compared with the results of California 8# model, its detection rate is increased by 187%,and its error detection rate is decreased by 46%,and its average detect time is shorted by 73%.6 tabs,4 figs,10 refs.