基于有趣地点压缩的时空轨迹聚类

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.

Clustering spatio-temporal trajectories based on compression of interesting places

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.