轨迹数据的非关系管理及相似性分析

Management and similarity analysis of trajectories with NoSQL database

针对高动态、半结构化的轨迹数据,充分利用文档型非关系数据库Mongo DB的特性,本文首先提出了一套分层、分区、分片的存储策略,设计了以整条轨迹为基本粒度的非关系组织模型,能够有效应对轨迹数据的海量性和动态性挑战。然后据此开展轨迹相似性分析的研究,提出了一种兼顾时间维和轨迹形状的轨迹相似性度量方法 DTWEUCLI,可计算长短不一且含有噪声的轨迹数据之间的相似性。最后基于轨迹的非关系存储和相似性计算,开展了轨迹簇生成的试验与分析,设计实现了基于轨迹相似性计算的轨迹聚类计算框架。基于3个轨迹数据集的试验表明,DTWEUCLI算法能够对多源轨迹数据集进行有效聚类,输出轨迹簇。

Based on the dynamic and semi-structured trajectory data,this paper makes full use of the characteristics of the documenttype non-relational database MongoDB,proposes a layered,partitioned and sliced storage strategy,and designs a non-relational organization with the whole trajectory as the basic unit. The model can effectively cope with the massive and dynamic challenges of trajectory data. On this basis,this paper studies the trajectory similarity calculation,and proposes a trajectory similarity measurement method DTWEUCLI that takes into account both time information and trajectory shapes,which can effectively calculate the similarity between trajectory data with different lengths and noises. Finally,based on trajectory-based non-relational storage and similarity calculation,this paper carries out the experiment and analysis of trajectory cluster. The trajectory cluster computing framework based on trajectory similarity calculation is proposed. Experiments based on three data sets show that the DTWEUCLI algorithm can effectively calculate the trajectory cluster of multi-source trajectory data sets and export the trajectory cluster.