【作者】 王强；

【导师】 王家耀；

【作者基本信息】 解放军信息工程大学 ， 地图制图学与地理信息工程， 2010， 博士

【摘要】 开放环境下的空间信息服务已成为空间信息共享与应用的主要手段,然而服务资源的多样性和不确定性、应用需求的动态性和复杂性使人们往往难以直接获得所需的信息。空间信息服务聚合作为一种新的服务模式,目的是按照合理的业务逻辑和语义约束来集成多个服务从而满足用户需求。这是一个复杂的、创造性的过程,迫切需要强有力的计算环境、一致的语义描述与理解、合理的知识表达以及严密的逻辑推理来支持。本文从计算和智能两个角度来实现空间信息服务聚合,将网格计算与人工智能中的技术与理论应用于空间信息服务的构建、描述、共享与集成。对空间信息服务聚合框架、高性能的空间信息服务、语义化的空间信息服务描述注册与发现、规则驱动的空间信息服务链自动规划等方面进行了深入研究,在服务聚合的理论和实践上取得了一些成果,主要的工作和创新点包括:1.结合地震堰塞湖的应急预警过程,深入分析了空间信息应用对服务聚合提出的新要求,指出服务语义不明确和服务效率低下是目前实现空间信息服务聚合面临的主要问题。提出将空间信息网格作为服务聚合的支撑环境,提供高效的分布并行计算能力;将以地理空间本体和描述逻辑为主的空间知识表达与推理框架作为服务聚合实现的理论基础和形式化手段,为其提供一致的空间信息资源表达和严密的逻辑推理能力。在此基础上,阐述了空间信息服务聚合的定义、框架以及实现的关键技术。2.研究了空间信息网格中的服务构建和优化的关键技术。基于WSRF规范重构现有的空间信息服务使其成为有状态的网格资源,实现网格技术与空间信息服务的融合;针对数据组织、传输和计算三个制约空间信息服务性能的关键环节,分别研究了基于聚类Hilbert R树和均衡数据划分策略的分布并行空间数据索引、基于GridFTP和网格资源管理机制的海量空间数据传输与获取、基于网格任务提交机制和扩展WPS接口的并行空间数据处理,并结合移动Agent技术研究了空间信息服务迁移计算模式,设计并实现了基于网格资源监控的服务迁移路径规划算法。3.研究了地理空间领域本体及其支持下的空间信息服务语义描述。依据空间知识的粒度和作用关系提出了由顶层本体、地理空间领域本体及应用本体构成的地理空间本体层次模型。针对目前缺少较完善的地理空间领域本体模型及实现方法的现状,基于ISO标准和OGC规范中的核心概念及关系提出了一个较完整的地理空间领域本体概念模型,研究通过UML到OWL元素的映射和语义精确化来构建地理空间领域本体的方法。从数据、功能、执行和服务质量四个方面详细分析了空间信息服务所蕴含的语义,研究了基于地理空间领域本体和服务本体描述语言OWL-S对空间信息服务进行描述的方法,为服务共享和聚合提供了有效的形式化和自动理解的基础。4.研究了语义支持的空间信息服务注册模型与描述逻辑支持的空间信息服务匹配方法。首先针对目前空间信息服务注册中心缺少语义支持的问题,扩展OGC推荐的ebRIM模型,将OWL描述的地理空间领域本体和OWL-S描述的空间信息服务映射到ebRIM模型中,实现了语义化的空间信息服务注册。针对服务匹配算法过于关注语义相似度计算,而忽视服务执行语义描述的问题,提出采用描述逻辑和Horn逻辑对空间信息服务的前提和效果进行形式化表达,从而实现基于逻辑推理的服务执行语义匹配。在此基础上提出多层次匹配模型,通过服务分类本体的粗略匹配、服务输入输出的语义匹配和的服务执行语义的精确推理匹配三个步骤,有效地减少了待匹配服务数量,提高了服务匹配准确性。5.研究了规则驱动的空间信息服务链建模与执行关键技术。首先提出空间信息服务聚合的四阶段实现方案,即服务链的规划、选择、服务实例化和执行。将人工智能规划中的反向搜索算法用于服务链的自动规划,制定了相应的控制规则和领域规则,将服务执行的逻辑语义同规则的前提和效果联系起来,挖掘空间信息领域特定的应用逻辑以及领域本体中的隐含知识进行服务聚合;针对服务规划形成的多条服务链,采用局部最优算法进行服务链的优化选择;利用OWL-S同WSDL的映射关系和服务匹配机制实现服务链同服务实例之间的绑定;基于OWL-S和BPEL两种服务链执行引擎来实现服务链的执行。6.基于论文研究成果,设计并实现了空间信息服务聚合平台,并结合堰塞湖灾害预警监测来构建典型应用案例,对论文所述模型、方法的可行性和有效性进行了验证。更多还原

【Abstract】 Spatial Information Services have become the most important way to acquire geospatial information on the open Internet environment. But the variety and uncertainty of the service resources, and the dynamic and complex user requirement blocked people from acquiring the information they needed efficiently and intelligently. The aggregation of Spatial Information Services is a new application schema for acquiring spatial information. Its goal is to composite several services by combining application logic and semantic constrains automatically to satisfy user requirement. In this complex and creative procedure, not only powerful computing environment should be guaranteed, but also the consistent semantic descriptions and understanding, reasonable knowledge representation and rigorous logic deduction should be supported.This thesis attempts to aggregate Spatial Information Services from computing and intelligent perspective. It introduces the theories in Artificial Intelligence and the techniques in Grid Computing to construct intelligent and efficient aggregation of Spatial Information Services. It focuses on the framework of service aggregation, high performance Spatial Information Services, semantic service description registry and discovery. It also studies the key issues on modeling of rule-driven geospatial service chaining. Its main achievement and innovation on models, methods and applications are described as follows.1.Using the subsequent disaster of earthquake scenario, the new requirements on aggregation of Spatial Information Services are analyzed. The semantic ambiguity and low efficiency are the main problems services aggregation is facing. In order to provide high performance distributed parallel computing power the Spatial Information Grid is suitable for environment. Geospatial ontology and description logic as the main knowledge representation frame are the theory base of service aggregation, which provide the consistent semantic representation and the ability of logic deduction. Then the definition and aggregation frame are introduced, the key technologies are discussed.2.The construction and performance improving methods are studied. Based on WSRF specification the current Spatial Information Services are wrapped into stateful resources. In order to provide high performance services, the Grid based geospatial data storage, process and transport are studied. The clustered Hilbert R-Tree spatial index and average divided data method are utilized for building parallel spatial data index. To improve the HTTP bottleneck on data transport, the GridFTP is introduced for spatial data transport and access. The spatial data process services are studied and the interface is extended for grid job submission. Based on Grid resource monitor mechanism and Mobile-Agent, the migratory geospatial information services are put forward.3.The geospatial ontology and semantic descriptions of Spatial Information Services are studied. Based on the granularity and relationship between geospatial knowledge concepts the top level, domain level and application level of geospatial ontology are discussed. Based on the current ISO and OGC standards, the geospatial domain ontology is constructed. The method of mapping elements from UML to OWL is put forward. The semantics of geospatial services are analyzed from four aspects which are data, function, execution and QoS. Based on the geospatial ontology and OWL-S, the geospatial services are formally described.4.Semantic supported service registry and Description Logic supported service matching are studied. First for the lack of semantic supporting, the ebRIM model is extended for registry which mapped the OWL and OWL-S into registry model. The Description Logic and Horn Logic are utilized for description and deduction of geospatial service precondition and postcondition in order to cover the weakness of execution semantic description. Based on the rules and deduction, improved the service execution match making. Above these methods, a multi-step geospatial service match making strategy is put forward.5.The model construction of rule based Spatial Information Services chain and execution are studied. In order to put service aggregation into realization, the four steps of aggregation are brought forward. First the backward search method in Artificial Intelligence is introduced for services planning. The domain rules and instructing rules are studied. The planning result is mapped to service DAG. Then the locally optimal method is introduced for picking up a service chain. Based on the relationship between OWL-S and WSDL, the service instances are binding to service chain. At last two service execution engines are discussed for service chain execution.6.Based on above research, the geospatial information service aggregation platform and sub modules are designed and implemented. With the application of Block lake risk monitor and warning, a use case has been discussed to validate the feasibility and validity of the models and methods.更多还原