基于动态分段和DBSCAN算法的交通事故黑点路段鉴别方法  

Identification method of traffic accident black spots based on dynamic segmentation and DBSCAN algorithm

作　　者：耿超[1,2] 彭余华 GENG Chao;PENG Yu-hua(School of Highway,Chang＇an University,Xi＇an 710064,Shaanxi,China;Anhui Expressway Company Limited,Hefei 230088,Anhui,China)

机构地区：[1]长安大学公路学院,陕西西安710064 [2]安徽皖通高速公路股份有限公司,安徽合肥230088

出　　处：《长安大学学报（自然科学版）》2018年第5期131-138,共8页Journal of Chang’an University(Natural Science Edition)

基　　金：云南省交通运输厅科技项目(云交科教(2016)56号)

摘　　要：为提高交通事故黑点鉴别的精度,选择合适的路段长度作为分析路段的基本单元,结合移动步长法将路段划分为几种不同的组合,兼顾交通事故次数和伤亡人数对交通事故进行当量化处理;采用累计频率法进行事故分析,分别得到各种组合下的事故相对集中路段,按路段相邻原则对其进行合并作为初选黑点路段。针对初选黑点路段,选取合适的邻域和阈值,采用基于密度的聚类(DBSCAN)算法进行聚类分析,以寻求长度较短且事故集中的真正事故黑点。为验证该鉴别方法的可靠性,对麻昭(麻柳湾—昭通)高速公路部分路段进行事故黑点鉴别。结果表明:在动态路段划分的基础上,采用当量事故数-累计频率法,初步鉴别出的事故多发路段长度占全线总长28.4%,发生当量事故次数占总事故次数的56.4%;通过移动步长法动态划分路段单元能够最大程度鉴别事故多发路段;采用DBSCAN算法对初选黑点路段进行再次排查,得到的事故黑点路段长度仅占全线总长10.8%,其发生当量事故次数占总事故次数的52.5%;与初选路段相比,其总长度下降了61%,总当量事故次数几乎不变。将动态分段和DBSCAN算法结合能够剔除初选路段中的非黑路段,识别出真正的黑点路段。该方法提高了公路交通事故黑点路段的鉴别精度,可为交通事故黑点的有效治理提供坚实的技术支撑。To improve identification accuracy of black spots in traffic accidents,an appropriate length of a road segment was selected as the basic unit for analysis and divided into several section combinations by the mobile step method.The mobile step method considered the number of traffic accidents and resulting casualties for a quantitative measure of traffic accidents.Theaccumulative frequency method was used to analyze the accidents,and obtain sections with relatively concentrated number of accidents under various combinations,then which were merged as the primary black spots according to the principle of adjacent sections.For primary black spots,appropriate neighborhood and threshold values were selected,and the density-based spatial clustering of applications with noise（DBSCAN）algorithm was used for cluster analysis to seek out the＂real＂black spots with short lengths and concentrated accident occurrences.To verify the reliability of the identification method,black spots were identified for various sections of the Mazhao Expressway.The results show that length of the accident-prone section identified by the accumulative frequency method accounts for 28.4% of the total length of the segment,and the number of equivalent accidents accounts for 56.4% of the total number of accidents,on the basis of dynamic segment division.Thus,dividing the road unit dynamically by the mobile step method can identify the black spots to a great extent.The lengths of the real black spots only account for10.8% of the total length,and the number of equivalent accidents accounts for 52.5% of the total number of accidents,When the DBSCAN algorithm is used to re-examine the obtained primary black spots.Thus the length of black spot sections have decreased by 61%,while the number of accidents remained almost the same.The combination of dynamic segmentation and the DBSCAN algorithm can be used to eliminate ＂non-black＂road sections in the primary road segment and identify the real black point segments. This method greatly improves the

关 键 词：交通工程 交通安全 事故黑点 累计频率法 当量事故数 DBSCAN算法 

分 类 号：U491[交通运输工程—交通运输规划与管理]