基于结构相似度的轨迹聚类算法

针对目前轨迹数据聚类直接以整条轨迹数据作为基本单元,导致聚类效果降低的问题,提出了基于结构相似度的轨迹聚类算法。算法引入轨迹结构的概念,并给出结构相似度计算函数来分析轨迹内外部特征。首先根据转角将轨迹划分成若干轨迹段,然后通过计算轨迹段的结构相似度来判断轨迹的匹配程度,进而完成轨迹聚类。真实数据的实验结果表明:该算法较其他同类算法分析轨迹更全面、效率更高;可以通过不同参数灵活调整特征的敏感度,聚类结果更具有实际意义。

Trajectory clustering algorithm based on structural similarity

For current trajectory clustering algorithms,most of them group full trajectories as basic units,and lead the low efficient results.Aiming at this problem,a trajectory clustering algorithm based on structural similarity was proposed.By introducing a new concept of trajectory structure and presenting structural similarity function,the internal and external features of trajectories were analyzed.The algorithm first partitioned trajectories into trajectory segments according to corner;then computed the matchin...