基于因子分析与聚类分析的交通事件自动检测算法融合

Automatic incident detection algorithms fusion method based on factor analysis and cluster analysis

针对目前通常只利用一种交通事件自动检测算法进行事件检测导致的效果不佳问题,在对单个交通事件自动检测算法产生漏警和误警的原因进行分析的基础上,设计了一套与之相对应的交通条件在线评价指标,并以因子分析与聚类分析为手段提出了一种基于多个交通事件自动检测基本算法的决策级融合方法。运用某特大城市快速路感应线圈实测数据进行验证的结果表明,在交通事件自动检测基本算法的误警率为0.5%左右、检测率为63.5%～66.1%的条件下,所提出方法的检测率和误警率分别达到了90.6%和0.0981%,明显优于对比方法的检测效果。

For the improvement of the effect of incident detection using the single automatic incident detection algorithm（AIDA）,a set of indexes used for traffic condition on-line evaluating was designed based on the analysis of the causes of failed alarms and false alarms generated by AIDAs used,and an AIDAs fusion method based on factor analysis and cluster analysis was developed.The proposed method was verified and compared using the data collected from the inductive loops on a metropolitan urban freeway.The results showed that under the condition of aimed false alarm rate 0.5% and the detection rates of the original AIDA 63.5 %~66.1%,the detection rate of proposed method is 90.6% and its false alarm rate is only 0.0981%,being obviously better than the contrast methods.