摘要
提出了一种基于支持向量机,利用行人交通流参数,实现城市轨道交通人行通道处交通状态识别的方法。采用FCM算法实现了4种行人交通状态的聚类分析与定义;建立SVM多类分类器模型。并分别采用线性可分和非线性两种SVM分类器以及多项式、高斯径向基、sigmoid等3种核函数,应用于行人通道交通状态识别中,进而通过实际采集数据集合,对其有效性进行对比分析。研究表明:设计的算法具有良好的识别性能,RBF核函数SVM模型的总体识别效果相比最好,正确率均在85%以上,说明行人交通参数在该核函数转化的高维空间具备良好的线性可分;线性可分对畅通状态识别效果相对最好,正确率为98%;多项式核函数对稳定状态识别效果相对最好,正确率为93%;sigmoid核函数的总体识别效果相比最稳定,正确率均在85%-92%。
A method based on support vector machine (SVM) was proposed, which used the pedestrian's traffic flow parameters to realize the pattern recognition of traffic condition of urban rail transit pedestrian passage. Clustering analysis and definition of four kinds of pedestrian traffic status was realized by FCM algorithm, and multiple SVM traffic status identification model was established. Linear and nonlinear SVM classifiers were adopted respectively, and polynomial kernel function, radically Gaussian kernel function and sigmoid kernel function were also adopted in the pattern recognition of pedestrian passage traffic condition. Furthermore, the comparison analysis on the effectiveness was carried out through the actually collected data collection. The results indicate that the proposed algorithm has good recognition performance. With the comparison of overall recognition performance, RBF kernel function SVM model is the best one and its correct rate is over 85%, which indicates that the pedestrian traffic parameters are well linearly separable in the high-dimensional space of the kernel function conversion. Linear separability has relatively best performance in smooth state recognition, and the correct rate is 98%. Polynomial kernel function has relatively best performance in steady state recognition, and the correct rate is 93%. With the comparison of overall recognition performance, sigmoid kernel function is relatively most stable one, and the correct rate ranges from 85% to 92%.
出处
《重庆交通大学学报(自然科学版)》
CAS
北大核心
2016年第3期134-140,共7页
Journal of Chongqing Jiaotong University(Natural Science)
基金
国家自然科学基金项目(50808021)
陕西省交通运输厅科技项目(10-07R)
关键词
交通运输工程
轨道交通
状态识别
支持向量机
traffic and transportation engineering
rail transit
pattern recognition
support vector machine (SVM)