基于多聚类结果融合的轨迹聚类方法  

Trajectory Clustering Method Based on Multi-clustering Results Merging

作　　者：李静[1] 张磊[1] 韩陈寿[1] 

机构地区：[1]中国矿业大学计算机学院,江苏徐州221116

出　　处：《微电子学与计算机》2011年第8期63-66,共4页Microelectronics & Computer

基　　金：国家自然科学基金项目(50674086);高等学校博士学科点专项科研基金项目(20060290508);江苏省博士后基金(0802023C)

摘　　要：针对轨迹聚类结果的不可靠性,提出一种基于多聚类结果融合的轨迹聚类方法MRMTC.对于多聚类器产生的多个聚类代表轨迹,提出了轨迹合并算法,实现了多个聚类代表轨迹的合并.代表轨迹合并算法以平均扫描线距离函数作为共识函数,通过共识函数对代表轨迹间的相似度进行比较,最后合并相似的代表轨迹.实验表明基于融合的轨迹聚类方法,可以获得比单一聚类更有效更稳定的聚类结果.In view of the unreliable of trajectory clustering results,a trajectory clustering method based on multi-clustering results merging（MRMTC） is proposed in this paper.For the representative trajectories of multi-clustering generated by clustering devices,a trajectory merging algorithm is proposed to merge them.The merging algorithm uses average scan line distance function as consensus function to compare similarities of representative trajectories,and then merges the similar representative trajectories.Finally,experiments results show that the proposed method MRMTC can produce more stable and effective clustering results.

关 键 词：轨迹 聚类 聚类融合 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]