基于网络核密度估计的交通事故黑点识别  

Identifying Urban Road Black Spots Based on Network Kernel Density

作　　者：罗叶 祁首铭 张希 廉冠 杨海华 LUO Ye;QI Shouming;ZHANG Xi;LIAN Guan;YANG Haihua(Shenzhen Urban Public Safety and Technology Institute,Shenzhen 518000,China;School of Civil and Environmental Engineering,Harbin Institute of Technology,Shenzhen 518055,China;School of Civil and Transportation Engineering,Guilin University of Electronic Technology,Guilin 541004,China;China Construction Infrastructure Co.,Ltd,Beijing 100029,China)

机构地区：[1]深圳市城市公共安全技术研究院有限公司,广东深圳518000 [2]哈尔滨工业大学土木与环境学院,广东深圳518055 [3]桂林电子科技大学建筑与交通工程学院,广西桂林541004 [4]中国建设基础设施有限公司,北京100029

出　　处：《大连交通大学学报》2023年第4期19-26,共8页Journal of Dalian Jiaotong University

基　　金：国家自然科学基金青年项目(52102410);广西青年科学基金(2020JJB170049);广西科技基地和人才专项(桂科AD19245021)。

摘　　要：为了对交通事故多发点进行分析,识别事故黑点并进行针对性治理,以深圳市龙华区2018—2020年3年间的1 105起道路交通事故数据及道路网络数据作为研究对象,采用高斯函数作为内核函数的道路网络核密度估计,通过K-means聚类进行路段密度等级划分。为进一步考虑路段之间的相关性,提高事故多发点识别精度,采用Local Moran′s I对事故多发点进行识别。对比平面核密度估计结果表明,所采用的基于网络核密度的Local Moran′s I的方法体现了事故多发点在路段及交叉口的分布特征。研究结论对识别事故黑点、进一步提高交通安全水平具有现实意义。Road traffic crashes cause considerable economic losses and endanger people's lives.1105 road traffic accident data between 2018 to 2020 and a road network in Longhua District,Shenzhen were used for the analysis and modeling.The Gaussian kernel was used to estimate the road network kernel density,and apply K-means clustering to classify the density of road sections.To improve the accuracy of identifying black-spot,the Local Moran's I method which consider the correlation of road sections was introduced.By comparing the results of plane kernel density estimation,the Local Moran's I method based on network kernel density used in this paper reflects the distribution characteristics of black spots in road sections and intersections.The research conclusions have practical significance for identifying black spots and further improving the level of traffic safety.

关 键 词：城市道路交通事故 网络核密度 事故黑点识别 K-MEANS聚类 Local Moran′s I 

分 类 号：U491.31[交通运输工程—交通运输规划与管理]