基于集成聚类的交叉口车辆行驶路径提取方法  

EXTRACTION METHOD OF VEHICLE DRIVING PATH AT INTERSECTION BASED ON THE ENSEMBLE CLUSTERING

作　　者：尹卓 许甜 陈阳舟[3] 卢佳程 Yin Zhuo;Xu Tian;Chen Yangzhou;Lu Jiacheng(Beijing Key Laboratory of Transportation Engineering,Beijing University of Technology,Beijing 100124,China;CCCC First Highway Consultants Co.,Ltd.,Xi’an 710075,Shaanxi,China;College of Artificial Intelligence and Automation,Beijing University of Technology,Beijing 100124,China)

机构地区：[1]北京工业大学北京市交通工程重点实验室,北京100124 [2]中交第一公路勘察设计研究院有限公司,陕西西安710075 [3]北京工业大学人工智能与自动化学院,北京100124

出　　处：《计算机应用与软件》2020年第7期180-187,193,共9页Computer Applications and Software

基　　金：国家自然科学基金项目(61573030);国家重点研发计划项目(2017YFC0803906,2017YFC0803900)。

摘　　要：交叉口不同方向车辆的行驶路径冲突会导致各种碰撞风险,因此利用交叉口的车辆轨迹数据提取该场景下的路径信息,对交叉口碰撞风险分析具有重要意义。提出一种基于集成聚类的路径信息提取方法,用于从交叉口车辆轨迹数据中获取交叉口路径信息。将车辆轨迹投射到网格上,并压缩其中大量的冗余点;采用集成聚类方法提取高质量的聚类,获取聚类后,基于图的拉普拉斯中心性,提取各个聚类的代表;采用深度优先搜索将聚类合并成路径。实验分析表明,该方法在多个数据集上展现出较强的鲁棒性,并且提高了路径提取精度。The path conflicts of vehicles in different directions at intersections will lead to various collision risks.Therefore,it is of great significance to extract the path information under the scene by using the vehicle trajectory data of intersections.This paper proposes a path information extraction method based on ensemble clustering to obtain intersection path information from the vehicle trajectory data.The vehicle trajectory was projected onto the image grid,and a large number of redundant points were compressed.The ensemble clustering method was used to extract high-quality clusters.After acquiring the clusters,the representative of each cluster was extracted based on the Laplacian centrality of the graph.Finally,the depth-first search was used to merge the clusters into paths.The experimental results show that our method has strong robustness in multiple datasets and improves the accuracy of path extraction.

关 键 词：轨迹聚类 轨迹特征提取 集成聚类 拉普拉斯中心性 图论 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]