为了发现交叉口各进口交通流量变化的细微差异,实现交叉口交通运行特性中潜在属性的挖掘,基于道路检测器数据,结合谱聚类算法和快速独立成分分析(Fast Independent Component Analysis,简称FastICA) 算法,提出了一种研究不同时间长度(1小时、1 天、1 周) 各进口交通运行特性的方法。以交叉口各进口道路检测器数据为研究对象,利用常规的数据处理方法得到短时交通运行特性,然后通过谱聚类算法挖掘出观测当天对向交通运行特性,采用FastICA 算法获得较长时间内各进口交通运行特性。最后,依据南昌市某交叉口数据进行了计算。结果表明,该交叉口西进口交通运行情况受不稳定交通流的影响,南、北进口交通流处于相对稳定的状态,东进口交通运行状态是整个交叉口交通运行状态的主要影响因素。基于此,上述方法的可行性得到了验证。
Abstract
In order to find the subtle differences in the traffic flow changes at each entrance of one intersection and realize the potential attribute mining of the intersection traffic characteristics, a method was proposed to study the traffic operation characteristics of each entrance of one intersection under different time lengths (one hour, one day and one week). This method based on the data of road detectors and combined spectral clustering algorithm with fast independent component analysis (FastICA) algorithm. Taking the road detector data of each entrance at the intersection as the research object, the conventional data processing method was used to obtain short-term traffic characteristics. Then, spectral clustering algorithm was used to mine the traffic characteristics of the two opposite entrances at the same day. The FastICA algorithm was used to get the long-period traffic characteristics of each entrance. Finally, the method was verified according to the data of one intersection in Nanchang. The results show that the traffic operation status of the west entrance of the intersection is affected by unstable traffic flow, and the traffic flows in the south and north entrances are in relatively stable states. The traffic operation status of east entrance is the main influencing factor of the traffic operation status of the entire intersection. Based on this, the feasibility of the method has been verified.
关键词
城市交通 /
道路检测器数据 /
不同时间长度 /
谱聚类 /
FastICA算法
Key words
urban traffic /
road detector data /
different time length /
spectral clustering /
FastICA algorithm
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 刘静,关伟. 交通流预测方法综述[J]. 公路交通科技,2004,21(3):82-85.
[2] 达庆东,段里仁. 交通流非参数回归模型[J]. 数理统计与管理,2003(4):42-47.
[3] 郑长江,李悦,王婷. 非常态下城市道路交通检测器布局优化方法[J]. 中国公路学报,2011,24(5):100-106.
[4] 林培群,徐建闽. 基于图论的路网交通检测器之布点[J]. 控制理论与应用,2010,27(12):1605-1611.
[5] 郭瑞军,林柏梁. 环形交叉口交织区车流运行特性的研究[J]. 交通运输系统工程与信息,2010,10(3):29-34.
[6] 刘东,韩凤春,陈永胜. 灯控路口非机动车交通流运行特性研究[J]. 中国人民公安大学学报(自然科学版),2004,10(2):89-92.
[7] 杜明洋. 信号交叉口非机动车和行人混合交通流运行特性[J]. 中国机械,2014(7):60-61.
[8] JIN S, QU X B, XU C, et al. Dynamic Characteristics of Traffic Flow with Consideration of Pedestrians′ Road-Crossing Behavior[J]. Physica A (Statistical Mechanics and Its Applications), 2013, 392(18): 3881-3890.
[9] MORIDPOUR S, ROSE G, SARVI M, et al. Effect of Surrounding Traffic Characteristics on Lane Changing Behavior[J]. Journal of Transportation Engineering, 2010, 136(11): 973-985.
[10] 李新叶,余晓晔. 适用于复杂结构的多路谱聚类算法的改进[J]. 北京工业大学报,2013,39(3):425-429.
[11] 何顺斌,刘杰. 基于谱聚类算法的岩体结构面产状优势分组[J]. 黄金科学技术,2017,25(4):46-51.
[12] 王灿灿,李丽娟. 改进谱聚类算法在多模型软测量中的应用[J]. 自动化仪表,2014,35(6):35-38.
[13] LUXBURG U. A Tutorial on Spectral Clustering[J]. Statistics and Computing, 2007, 17(4): 395-416.
[14] 蔡晓妍,戴冠中,杨黎斌. 谱聚类算法综述[J]. 计算机科学,2008,35(7):14-18.
[15] 卜德云,张道强. 自适应谱聚类算法研究[J]. 山东大学学报(工学版),2009,39(5):22-26.
[16] 徐天顺. 谱聚类算法研究[J]. 电脑知识与技术,2012,8(16):3948-3950.
[17] 杨金成,张南. 独立成分分析技术综述[J]. 舰船科学技术,2007,29(2):83-86.
[18] 郭武,朱长仁,王润生. 一种改进的FastICA算法及其应用[J]. 计算机应用,2008,28(4):960-962.
[19] 张君生,刘瑞安,王磊. 基于FastICA的工频干扰消除算法[J]. 天津师范大学学报(自然科学版),2013,33(2):38-42.
[20] 王建雄,张立民,钟兆根. 基于FastICA算法的盲源分离[J]. 计算机技术与发展,2011,21(12):93-96.
[21] 李睿,黄旭. 一种改进的FastICA信号盲分离算法[J]. 通信对抗,2010(2):12-15.
基金
国家自然科学基金重点项目(51338008);国家自然科学基金面上项目(51378036);西宁市科技项目
(2015-SF-A5) 子项目(2003617195);华东交通大学博士启动基金项目(2003417029)