【作者】 谢炯；

【导师】 刘南；

【作者基本信息】 浙江大学 ， 地图学与地理信息系统， 2005， 博士

【摘要】 下一代GIS必须具备对海量时空数据进行有效的表达、管理与分析能力,而时空数据模型是解决上述问题的基础和关键。由于时空变化语义的复杂性、时间维表达的特殊性、动态多维扩展后技术实现的繁难性以及海量空间信息考虑时变因素后的超海量性,致使目前仍无普遍接受的时空数据模型;GIS基础平台研制和应用系统开发主体仍沿用传统空间数据模型与建模方法,已难以满足GIS对时空应用发展的需求,尤其是对海量时空数据管理与分析的需求。本研究以面向实体和多尺度时空过程统一集成描述的无缝时空思想为指导,采用基于地理特征域、时空场域、事件域以及关联域相综合的多重表达方法,按时空语义建模、时空数据逻辑建模和时空数据物理建模三个层次对无缝时空的多域集成时空数据模型(SMDI-STDM)进行了理论、技术到实践的综合探讨。模型在更基础层面上总结了不同地理事物和现象的时空变化特点,力图更为完整、有效地表达时空变化语义,设计相应的数据组织结构,并在解决海量时空数据的高效组织与存储问题上力争有所突破。主要研究内容包括:(1)综述。从空间语义、专题语义、时间语义及时空语义出发,构建了时空表达的一般性概念框架,讨论了国内外时空建模现状、发展趋势和面临的问题。(2) SMDI-STDM语义模型。提出了面向实体时空变化与多尺度时空过程集成统一描述的时空变化刻画的三层次论,并以此为主体命名为“无缝时空”思想。以无缝时空思想为指导,对时空语义的基本表达对象——①地理特征与时空场;②时空链与时空图;③事件与事件链作了分类讨论。最后,通过关联域集成地理特征域、时空场域和事件域,提出了基于时空多重表达方法的SMDI-STDM模型,并通过多域集成表达,对复杂时空变化现象进行了时空语义建模,包括“特征-场”联合建模和面向时空变化刻画三个层次的综合建模。(3) SMDI-STDM逻辑模型。以实现无缝时空的多域集成表达为目标,设计了两层架构的逻辑组织模型:①底层面向对象-关系型数据库进行了时空综合扩展,对空间数据类型与操作、时态数据类型与操作、时空数据类型与操作,时空关系与时空关系代数操作,以及时空关联关系作了形式化定义,并对地理特征的时空五域结构、时空场的四维时空块结构、多域关联结构、特征-场的数据组织结构,以及面向海量时空数据的时空图库集结构进行了讨论。②中间层则以面向对象方法实现了统一的逻辑数据组织,对时空对象、时空数据集、时空变化组织和时空实体关系与规则进行了类的层次关系设计。(4) SMDI-STDM物理模型。针对当前海量时空数据管理的艰难性和紧迫性,将GIS时空数据特点与当前底层数据库技术有机结合,提出了基于时空分区与时空聚簇思想和方法的海量时空数据组织与存储新模式。时空分区方法实现时空非聚集实体的分块分磁盘并行存储,时空聚簇方法则进一步在时空分区内部实现时空邻近与物理存储位置也邻近的映射关系。(5)原型、测试与应用。讨论了原型系统(GeoST)的基本架构、功能模块划分和开发策略,给出了多域集成表达系列原型,并采用了2～60GB数据量的单表对时空分区与时空聚簇机制进行了分级效率测试。最后,给出了应用实例。研究、测试与应用表明,SMDI-STDM模型的设计思想是有效的、合理可行的。面向实体和多尺度时空过程统一集成建模的设计思想和方法能更真实地表达地理事物和现象的本质特性,也更能被人们所接受;针对时空语义的特殊复杂性,采用多域集成多重表达方法是必要的,原型实例进一步说明了方法的可实现性;海量时空数据的测试充分验证了时空分区与时空聚簇机制的有效性,从而为当前迫切需要解决的海量时空数据高效管理问题提供了切实可行的解决方案。更多还原

【Abstract】 A spatio-temporal data model is the critical basis of the abilities to effectively represent, manage and analyze massive spatio-temporal data,which is essential to next generation GIS. None of the existing spatio-temporal data models has been widely accepted caused by the complexity of spatio-temporal change semantics,the extraordinarity of the representation of time dimension,the complexities in the technical implementation of dynamic multi-dimensional extension and overwhelming mass data generated from adding time dimension to the massive spatial data.The ongoing researches and development of GIS base platform and applications may still employ traditional spatial data models and modeling methods,which can no longer satisfy the requirement of spatio-temporal applications in GIS, especially for the management and analysis of massive spatio-temporal data.Based on spatio-temporal semantic modeling,logical modeling and physical modeling methodology,this paper presents the theory,techniques and application of seamless temporal-space oriented multi-domain integrated spatio-temporal data model(SMDI-STDM), which is oriented to integrative description for entity changes and processes-the principles of seamless temporal-space,and uses the multi-representation approach integrating feature domain,spatio-temporal field domain,event domain and relation domain.The spatio-temporal changes of many geography entities and phenomenon are summarized and a corresponding data structure is designed to achieve a more complete and effective representation for spatio-temporal change semantics.The model also develops a new method for efficiently organizing and storing massive spatio-temporal data.The main content includes as follows:(1) Literature review.Typical spatio-temporal data models are reviewed and examined which is guided by a general conceptual framework of spatio-temporal representation constructed based on spatial semantics,attribute semantics,temporal semantics and spatio-temporal semantics.(2) SMDI-STDM semantic model.The seamless temporal-space concept mainly consists of the three-level-description of the spatio-temporal changes oriented to a integrative description for the entity changes and the spatio-temporal processes.Guided by the seamless temporal-space concept,basic objects for the representation of the spatio-temporal semantics are examined,which includeds feature and spatio-temporal field,spaito-temporal chain and spatio-temporal graph,event and event chain.Then, by utilizing relation domain integrating feature domain,spatio-temporal field domain and event domain,the SMDI-STDM model based on the multi-representation approach is presented.Using the multi-domain integrated representation method,the feature-field modeling and modeling for three-levels-description of spatio-temporal changes are further discussed.(3) SMDI-STDM logic model.To implement seamless temporal-space oriented multi-domain integrated representation,a logic organization model with two layers structure is designed.①On the lowest,object-relational database level, spaito-temporal extension is constructed.Spatial data types and operations,temporal data types and operations,spatio-temporal data types and operations,spatio-temporal relation and spafio-temporal algebra operations are formally defined as well as for spatio-temporal relationship.Features having spatio-temporal 5-domain structure, spatio-temporal fields in 4D spatio-temporal block structure,multi-domain relation structure,feature-field structure and spatio-temporal lib-dataset structure oriented to massive spatio-temporal data are examined.②On the middle layer,unified organization of logical data is implemented using an object oriented approach. Classes for spatio-temporal object,spatio-temporal dataset,spatio-temporal change organizer and spatio-temporal entity relationship and rule are designed.(4) SMDI-STDM physical model.In order to tackle the current difficulties of massive spatio-temporal data management,a new mode to organize and store massive spatio-temporal data based on spatio-temporal partition and spatio-temporal cluster are developed by the integration of GIS spatio-temporal data feature with current database technology.The spatio-temporal partition allows spatio-temporal unclustering entities to be stored in parallel structure at different blocks or disks. Further the spatio-temporal clustering presents the capacity of mapping entities neighboring in temporal space to their positions neighboring in physical storage.(5) Prototypes,tests and applications.A basic framework,function modules and the development strategy of Prototype GeoST are described.Series prototypes of multi-domain integrated representation are introduced.The efficiency of spatio-temporal partition and clustering method on single table from 2GB to 60 GB is tested.Finally,applications are presented.The experiments and applications show that the design of SMDI-STDM is effective, efficient,logical and practical.The modeling method based on integrative description of entity change and process could more realistically capture the essential features of geography entities and phenomenon,and may also be more acceptable.To solve the complexity of spatio-temporal change semantics,the multi-representation approach might be necessary and prototypes further show the practicability of the theory.The experiments with massive spatio-temporal data reveal that spatio-temporal partition and clustering method are efficient and logical,based on which a practical solution to effectively manage the massive spatio-temporal data could be marketed to users.更多还原