高速公路换道行为的频谱结构识别及分类  

Spectrum structure identification and classification of freeway lane changing behavior

作　　者：齐龙 张敏 徐婷[3] 陈亦新 QI Long;ZHANG Min;XU Ting;CHEN Yixin(School of Automotive Engineering,Shandong Jiaotong University,Jinan 250357,China;Taian Traffic and Transportation Bureau,Taian 271000,Shandong,China;College of Transportation Engineering,Chang'an University,Xi'an 710064,China)

机构地区：[1]山东交通学院汽车工程学院,济南250357 [2]泰安市交通运输局,山东泰安271000 [3]长安大学运输工程学院,西安710064

出　　处：《安全与环境学报》2023年第10期3520-3527,共8页Journal of Safety and Environment

基　　金：国家自然科学基金项目(51878066);陕西省自然科学基础研究计划项目(2021JQ-276);山东交通学院科研项目(Z201901)。

摘　　要：为提高高速公路出口换道车辆与主流车辆之间安全性,以驾驶行为功率谱分析为基础,进行换道识别及危险特征分类。首先,采集高速公路出口车辆车头时距、速度、加速度特征数据,从车辆行驶横向、纵向运行方向计算碰撞时间(Time-To-Collision,TTC),以得到危险冲突与一般冲突发生概率。其次,基于特征参数划分换道过程中安全状态,根据换道车辆运动学规律,对换道行为数据抽样,同时进行频谱结构分析,在定阶为10的频谱下构建行为功率谱模型,得到碰撞安全阈值。此外,通过相关性分析得到一般冲突和严重冲突的特征量,并与实际换道数据进行匹配。最后,利用K-means聚类分析对危险行为进行分类。研究结果显示:K-means聚类分析算法的精度较高,能有效辅助交通管制措施。研究成果可为高速公路换道行为研究和高速公路出口复杂交通变化研究提供理论基础。To improve the safety between freeway exit lanechanging vehicles and mainstream vehicles,based on the power spectrum analysis of driving behavior,lane-changing recognition and clustering hazard feature classification under the spectral structure analysis are carried out.Firstly,the UAV is used to collect the traffic flow data at the exit of the expressway,and the relevant data analysis software is used to extract the characteristic data of headway,speed,and acceleration.The collision time value is calculated from the transverse and longitudinal directions of the vehicle,and the probability of dangerous conflict and general conflict is obtained.Then,to use the spectrum structure to achieve the purpose of identification and determine the analysis accuracy order,the safety state in the process of lane changing is divided based on the characteristic parameters.Combined with the kinematic law of lane-changing vehicles,the lane-changing behavior data is sampled,and the spectrum structure is analyzed at the same time.The final spectrum structure is the best when the order is 10,and the behavior power spectrum model is constructed to obtain the safety threshold.After using the spectrum structure to identify the lane changing features,through the comparison of the previously calculated horizontal and vertical dangerous conflict and general conflict TTC data,the method is to conduct correlation analysis between the TTC calculation data and the data obtained from the power spectrum estimation and analysis of the characteristic quantity,which shows that there is an obvious correlation.Finally,because K means clustering has obvious adaptability in classifying lane change data,K means clustering analysis is used to classify dangerous behaviors.The research conclusion proves that the lane-changing behavior at the exit of the expressway can be identified by matching the eigenvalues of spectral structure,and the behavior can be classified by K means clustering analysis.

关 键 词：安全系统学 交通安全 换道行为 频谱结构 K-means聚类分析 

分 类 号：X951[环境科学与工程—安全科学]