高速公路出入口运动车辆轨迹分层聚类算法  

Hierarchical clustering algorithm of moving vehicle trajectories in entrances and exits freeway

作　　者：孙宗元[1] 方守恩[1] 

机构地区：[1]同济大学道路与交通工程教育部重点试验室,上海201804

出　　处：《吉林大学学报（工学版）》2017年第6期1696-1702,共7页Journal of Jilin University:Engineering and Technology Edition

基　　金："863"国家高技术研究发展计划项目(2013AA12A206)

摘　　要：为了提高对高速公路出入口车辆运动行为的理解和分析水平,根据出入口车辆运动轨迹的时空特征,提出了一种运动轨迹层次聚类算法。结合出入口轨迹方向一致、长短不一的特点,提出采用改进Hausdorff距离来衡量轨迹间的相似性。建立了改进模糊C均值轨迹分层聚类算法,首先根据轨迹的空间几何位置进行路径聚类,然后根据车辆的速度信息对已有路径聚类进一步聚类获得具有时空区分度的最终结果。真实高速公路出入口的试验结果表明:本文提出的轨迹聚类算法对于场景固定运动行为模式不仅具有较强的适用性,而且能够保障聚类结果的准确性和可靠性。In order to improve the understanding and analysis of motion patterns of vehicles, a hierarchical trajectory clustering algorithm is developed according to spatial and temporal characteristics of the vehicle trajectories in the entrances and exits of freeway. In view of the vehicle trajectories are different in length, but in the same direction, the improved Hausdorff distance was proposed and applied to measure the similarity of trajectories. The improved fuzzy C-means hierarchical clustering algorithm of trajectories was further established, in which trajectories were first clustered into different paths according to the spatial geometric position of the trajectories, and then trajectories belonging to the same path were further clustered according to the vehicle speed to obtain the final results with spatial and temporal degree. Experiments in the entrances and exits of freeway were carried out. The results confirm that the proposed trajectory clustering algorithm not only has strong adaptability to the inherent motion pattern of the scene, but also ensures the accuracy and reliability of the clustering results.

关 键 词：交通运输系统工程 高速公路出入口 轨迹分析 改进HAUSDORFF距离 聚类算法 

分 类 号：U491[交通运输工程—交通运输规划与管理]