摘要
针对利用最小包围盒(MBB)压缩的移动物体时空轨迹,为了能对其进行有效地聚类,提出了一个基于盒内数据点密度的轨迹间相似性度量公式.首先,把两条轨迹的相似性度量转化为两条轨迹上有时间交叠的MBB之间的相似性度量,这在很大程度上减少了数据存储量.其次,分析两条轨迹上有时间交叠的MBB之间影响相似性的因素:时间持续、空间距离和盒内数据点的密度.剖析这3个因素对轨迹相似性的影响作用,提出了利用MBB压缩的移动物体时空轨迹相似性度量公式.实验证明采用本公式对移动物体时空轨迹进行聚类,可以提高聚类结果有效性指标Dunn的值.
A similarity measurement formula is proposed based on the density of data points inside the boxes in order to effectively cluster the spatio-temporal trajectories of moving objects which are compressed into minimum bounding boxes (MBBs). The similarity measurement of the raw trajectories is translated into the similarity measurement of the MBB sequences with time overlapping in two trajectories firstly, which reduces data storage volume to a great extent. Then some factors affecting the similarity of MBB sequences are analyzed, including the time duration, the space distance and the density of data points inside the boxes. Through analyzing the influence of the three factors on the trajectory similarity, a formula of the spatio-temporal trajectories compressed into MBB is compressed into MBB is proposed. Experiments show that the formula can improve the value of validity index Dunn when it is used to cluster the spatio-temporal trajectories of moving objects.
出处
《信息与控制》
CSCD
北大核心
2012年第1期63-68,共6页
Information and Control
基金
轨道交通控制与安全国家重点实验室资助项目(RCS2009ZT007)
北京市科委资助项目(Z090506006309011)
国家科技支撑计划资助项目(2009BAG12A10)
关键词
时空数据挖掘
移动物体轨迹
轨迹聚类
轨迹相似性度量
spatio-temporaldata mining
moving object trajectory
trajectory clustering
trajectory similarity measurement