大规模轨迹数据的地理空间关联解译及分析挖掘研究

With the maturity of GPS, telecommunication and Web technology, and the wide adoption of GPS device, it is now relatively easy and cost-efficient to collect movement data of moving objects (like human beings, vehicles and wild animals) in the form of trajectory. As a result, massive amounts of trajectory data have been created in various areas. As trajectories always occur in the geographical space and carry richful information/knowledge,they have deep implications in many applications, e.g., location-based service, traffic control, city planning and animal protection. Consequently, more and more reseachers,particularly from the fields of spatial information and database,are attracted in the study on trajectory data, which resulted in abundant research results. However, the spatio-temporal relationships between trajectories and geographical features and their analytical applications have not been investigated too much, and therefore introduce great exploring potential. By linking trajectory with the contextual geogrphical information, this proposal will conduct comprehensive studies on geographically associated interpretation, analysis and mining for large trajectory dataset.The main work of this proposal includes: 1) Develop a spatio-temprally associated trajectory model that orientes to geographical information, and consequently, a trajectory of point sequence could be evolved into a semantic object with geographical interpretation; 2) Investigate the methods for the extraction of Stop/Move objects and the computation of spatio-tempral trajectory-feature associations, and explore the methods for building the spatio-temporally associated trajectory database; 3) Conduct trajectory query processing and analysis against the spatio-temporally associated trajectory database from three different aspects, i.e., trajectory->feature, feature->trajectory and trajectory->trajectory,and therefore, trajectory analysis of spatio-temporal topological relationship, spatio-temporal accessing relationship and spatio-temporal correlated relationship will be investigated; 4) Based on the spatio-temporal trajectory-feature associations, study two trajectory-related classical data mining problems, i.e., the computation of trajectory similarity and the mining of frequent accessing sequences. The research of this proposal will enrich the analytical methods in the field of trajectory study from the pointview of geographical association, and will lay a solid technical foundation for building the geographically oriented trajectory applications.

轨迹发生于地理空间之中，蕴含丰富的信息与知识，已成为数据库与空间信息等众多领域的研究热点，并取得了丰硕的研究成果。针对已有研究在轨迹-要素间复杂时空关联关系的建模与分析方面的不足，课题将面向丰富的地理空间上下文信息，开展针对大规模轨迹数据的地理空间关联解译及分析挖掘研究，主要工作包括：1）研究时空关联于地理空间要素的轨迹模型，将轨迹从点串数据升华为语义对象；2）研究Stop/Move对象的提取算法和轨迹-要素时空关联语义的计算方法，以及时空关联轨迹库的构建技术；3）面向时空关联轨迹库，从轨迹->要素、要素->轨迹、轨迹->轨迹三方面开展涵盖时空拓扑特征、时空访问特征、时空相关特征的轨迹数据查询分析；4）围绕轨迹-要素时空关联语义，开展轨迹相似性计算与频繁访问序列挖掘研究。通过课题研究，从地理空间关联角度丰富轨迹数据的分析手段，为建立大规模轨迹数据的应用分析系统奠定技术基础。

为了从海量轨迹数据中提取隐藏的信息与知识，学者们在组织模型、查询方法、分析技术、挖掘算法与行业应用等多方面开展了大量工作，取得了丰硕的研究成果。总体而言，已有工作在轨迹-地理空间要素间复杂时空关联关系的语义建模和面向时空关联轨迹的分析挖掘两方面的研究还比较薄弱。为此，我们面向丰富的地理空间上下文信息，建立轨迹-地理空间要素时空关联的语义轨迹模型，不仅可以显著提升轨迹模型的可理解性，而且能够开展更为明确、更具语义的查询、分析与挖掘任务，其结果也将更易于为用户所解译。.在轨迹数据地理关联建模方面，我们分别从轨迹-线、轨迹-面和轨迹-要素三个方向进行了研究，通过考虑轨迹在移动过程中的关键点，如起止点、停留点、相交点等，分别提出了基于关键点的轨迹-有向拓扑过程模型、基于关键点的轨迹-面拓扑过程模型，以及面向关键点的轨迹-要素模型，从而将轨迹的点串序列上升为语义编码序列，有效支撑了基于地理关联轨迹的分析挖掘。.在地理关联轨迹计算方面，主要从停留提取、存储模式和索引构建三个方向进行了研究。提出了基于序列聚类和K函数扩展的轨迹停留听取方法，设计了面向地理关联轨迹的组织模式，以及基于非关系数据库的存储技术，研制了基于扁平化R树的关联轨迹查询处理方法。.在轨迹关联分析挖掘方面，主要从轨迹地理关联解译和道路提取两方面进行了研究。提出了基于地理关联轨迹模型的新型分析技术，支持位置-时间、位置-顺序和位置-关系三类轨迹时空模式分析，发明了基于轨迹关联和地图匹配的道路发现方法，不仅可以检测与提取新建道路，而且能够发现与纠正数字道路的拓扑错误。.经过四年研究，项目共发表文章11篇，申请1项专利，其中第1标注的11篇，包括4篇SCI论文，5篇EI论文，1篇会议论文，2篇其他文章，项目建立较为系统的地理关联轨迹模型及其解译方法，研究成果符合预期。