基于有趣地点压缩的时空轨迹聚类  

Clustering spatio-temporal trajectories based on compression of interesting places

作　　者：赵秀丽[1,2] 徐维祥[1] 

机构地区：[1]北京交通大学交通运输学院,北京100044 [2]山东轻工业学院商学院,济南250353

出　　处：《北京交通大学学报》2011年第3期53-57,61,共6页JOURNAL OF BEIJING JIAOTONG UNIVERSITY

基　　金：国家科技支撑计划项目资助(2009BAG12A10);轨道交通控制与安全国家重点实验室支撑项目资助(RCS2009ZT007);北京市科委计划项目资助(Z090506006309011)

摘　　要：研究移动物体时空轨迹局部关键地点时空相似的聚类问题.根据移动物体的运动状态提取轨迹中的有趣地点,利用最小包围盒技术对这些有趣地点进行描述,得到基于有趣地点压缩的轨迹表示形式;然后给出一个时空属性相结合的相似性度量公式,对压缩表示的轨迹进行相似性度量;基于这个相似性度量公式对轨迹进行聚类,聚类方法采用层次聚类法.实验结果表明,本文提出的方法能有效地对移动物体时空轨迹进行聚类,由于采用了增量式的轨迹压缩方法,不仅提高了聚类的速度,而且还实现了增量式的轨迹聚类.Discovering similar trajectories according to proximity in time and space can greatly affect many fields such as animal migration,weather forecasting,and the personal and vehicular mobile patterns in urban transportation.This paper researches clustering trajectories left behind moving objects according to the spatio-temporal similarity of local interesting places on the trajectories.Firstly,these interesting places on the trajectories are extracted and turned to Minimum Bounding Boxes（MBB）,thus the original trajectories can be expressed by the smaller and less complex primitives（MBB） that are batter suited to storage and computation;and then a similarity measure formula is proposed with a combination of temporal and spatial properties of the compressed trajectory.Finally a hierarchical clustering experiment is performed in order to test the performance of the new similarity measure.The experimental results show that the proposed method not only can effectively cluster moving object trajectories,but also enables the clustering trajectory incrementally.

关 键 词：轨迹聚类 轨迹压缩 轨迹相似性度量 增量聚类 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] TP181[自动化与计算机技术—计算机科学与技术]