基于机器学习的地理信息链接方法研究

发布时间：2018-07-04 13:16

本文选题：地理信息系统 + 地理信息链接　；参考：《华北电力大学(北京)》2017年硕士论文

【摘要】：GIS是最近二十多年来新兴的一门集地理信息科学、计算机科学、测绘科学和统计科学等于一体的一门综合性学科,是用于输入、存储、查询、分析和展示地理数据的信息系统,以便及时解决地理信息处理中复杂的规划和管理问题。目前,GIS技术已经广泛的应用在各种的领域当中,GIS和应用模型的集成以及GIS智能化是拓宽GIS应用领域的关键。在地理信息系统中,不同地理信息源的内容之间的多样性、异构性等对地理信息实体描述的准确性、完整性等有很大的差异,这对实现地理信息共享以及促进地理信息技术的发展都产生了阻碍。不同地理信息源的地理信息实体之间链接的精确性对解决地理信息异构性,促进地理信息检索服务的准确性和地理信息集成等问题具有十分重要的意义。目前,大多数的研究工作基于语义关系、信息内容、上下文信息等来计算地理信息实体的相似性,忽略了地理空间关系和空间拓扑结构的作用。本文提出了一种基于空间关系、实体名称和实体类别等多特征的方式,同时结合语义和机器学习的方法实现的地理信息链接的半自动化方式。首先,分别从三个地理信息源:OpenStreetMap、Wikimapia、Google places抽取地理信息,抽取的地理信息主要针对美国洛杉矶和英国伦敦两个地区的城区建筑。其次,分析抽取地理信息数据的特点构建地理信息本体,通过地理信息源数据与地理信息本体映射,实现地理数据的一体化。最后,分别讨论融合分类算法支持向量机、K近邻方法的链接方法,同时与Samal等人提出的链接方法进行对比,多角度综合验证本文提出方法的准确性,为地理信息集成奠定了良好的基础。
[Abstract]:GIs is an integrated subject that integrates geographic information science, computer science, mapping science and statistical science. It is an information system for input, storage, query, analysis and display of geographic data. In order to solve the complex planning and management problems in geographic information processing in time. At present, GIS technology has been widely used in various fields of GIS and application model integration and GIS intelligence is the key to broaden the application of GIS. In GIS, the diversity and heterogeneity of the contents of different geographic information sources have great differences in the accuracy and completeness of the description of geographic information entities. This hinders the realization of geographic information sharing and the development of geographic information technology. The accuracy of the links between geographic information entities from different geographic information sources is of great significance in solving the problems of geographic information heterogeneity, promoting the accuracy of geographic information retrieval services and geographic information integration. At present, most of the research work based on semantic relations, information content, context information to calculate the similarity of geographic information entities, ignoring the role of geospatial relationships and spatial topology. In this paper, a semi-automatic method of geographic information link based on spatial relationship, entity name and entity category is proposed, which combines semantic and machine learning methods. First of all, three geographic information sources: OpenStreetMap WikimapiaGoogle places are extracted, which are mainly aimed at the urban buildings in Los Angeles and London. Secondly, the characteristics of extracting geographic information data are analyzed to construct geographical information ontology, and the integration of geographical data is realized by mapping geographical information source data with geographic information ontology. Finally, the link method of the fusion classification algorithm support vector machine / K-nearest neighbor method is discussed, and compared with the link method proposed by Samal et al., the accuracy of the proposed method is verified by multi-angle synthesis. It lays a good foundation for geographic information integration.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P208;TP181

【参考文献】