摘要
高速公路事件检测是交通信息工程及控制学科中的一项重要研究课题,以高速公路交通流的特点为研究对象,提出了一种基于支持向量机(SVM)的高速公路事件检测算法.根据支持向量机的基本原理,分别设计了基于线性不可分SVM、齐次多项式核函数、高斯径向基核函数、双曲线正切核函数等不同核函数的事件检测算法.仿真结果表明:针对不同的交通流状况,选择合适的SVM模型和核函数,得到的检测结果与经典的加利福尼亚算法相比,检测效率高,性能指标好,具有较高的实际应用价值.
Freeway Incident Detection （FID） has been studied as an important research subject in Traffic Information Engineering and Control . This paper , based on the freeway traffic flow characteristics ,presents new algorithms for FID with the support vector machine （SVM ） .According to the principal of SVM ,four different simulation experiments are carried out based on the linearly non-separable SVM ,the homogeneous polynomial kernel function ,the radial basis kernel function and the sigmoid kernel function respectively .The testing results obtained by the new algorithms were compared with those by the California algorithm .The simulation results show that choosing the right SVM model and kernel function according to different traffic flow characteristics can achieve better performance than the California algorithm does .The new algorithms prove of greater practical value .
出处
《西安工业大学学报》
CAS
2014年第9期726-731,共6页
Journal of Xi’an Technological University
基金
中央高校基本科研业务费专项资金资助(0009-2014G1241046
0009-2014G1321034
0009-2014G1241044)
关键词
高速公路
事件检测
支持向量机
参数优化
核函数
freeway
incident detection
support vector machine
parameter optimization
kemel function
