空间关系模型生成方法研究
发布时间：2018-07-28 21:52
【摘要】：随着信息技术的发展,信息量骤然增大,信息的价值密度降低,庞大的信息量看似能够给人们提供更多想要的信息,事实上却是使人更难使用信息。因此,各种大数据技术应运而生,旨在对海量的数据进行专业化处理,通过对数据的获取、存储、管理和分析,为人们提供真正需要的信息。空间信息作为海量信息的一种,被应用于人类生活的各个方面。无论是在互联网中新闻的推送、电子商务中广告的精准投放,还是导航和交通路况中对信息的分析,还是灾难后的政府救援,抑或是城市和企业发展中建筑的选址,都离不开大数据支持和对地理空间信息的分析。因此,对海量数据中空间关系的分析处理显得尤为重要。空间关系的研究已经有20多年的历史,空间关系被广泛地应用于空间推理和空间分析中,国内外已有不少学者做出了许多研究成果。现有的定性空间关系模型大多基于健全的代数系统,使用先验知识人工进行建模,现有简单空间关系模型不能全面地表达实际应用中所需要的空间关系;现有复杂空间关系模型仅能针对特定的复杂空间关系进行识别,不具有通用性。用基于健全的代数系统的方法对大数据中空间关系人工构建空间关系模型,需要对空间对象的先验知识进行分析,需要考虑的元素很多,导致形式化表示系统极其复杂,甚至难以形式化表示,不能满足大数据时代中对空间关系表达的需求。为了解决现有空间关系模型中存在的以上问题,我们对空间关系推理方法进行了研究,本文提出了将定性空间推理与机器学习相结合生成空间关系模型的方法。本文的方法能够在不需要先验知识的情况下生成空间关系模型,具有很好的通用性。本文的主要研究内容如下:(1)本文首先对空间推理中的空间关系这一研究热点的在当今时代的研究背景、研究意义以及目前的研究发展现状进行了阐述,然后介绍了本文的研究内容和全文的组织结构。(2)介绍了空间关系的基本概念和机器学习的相关知识,并对MLNB算法进行了描述。(3)通过对现有空间关系模型中存在的问题的分析,提出了生成空间关系模型的方法,介绍了该方法的流程。提出了空间关系的通用特征集合,详细介绍了对空间对象的区域分解,子区域和扩展区域的划分。并在此基础上,提出了特征分级策略,将子区域分组并组合,以减少处理的特征数目。(4)将本文提出的生成空间关系模型的方法应用于多个数据集,通过对比应用于不同数据集的结果分析了本文方法的优点。将本文方法应用于中文文本的空间关系识别中,取得了较好的效果。(5)对全文的内容进行总结,并提出了对进一步工作的展望。
[Abstract]:With the development of information technology, the amount of information increases suddenly, and the value density of information decreases. The huge amount of information seems to provide more information to people, but in fact, it makes it more difficult for people to use information. Therefore, a variety of big data technologies emerge as the times require, aiming at the specialized processing of massive data. Through the acquisition, storage, management and analysis of the data, it provides people with the information they really need. As a kind of massive information, spatial information is applied to every aspect of human life. Whether it is the push of news on the Internet, the precise delivery of advertisements in e-commerce, the analysis of information in navigation and traffic conditions, the rescue of governments after disasters, or the location of buildings in the development of cities and enterprises, Can not do without big data support and analysis of geospatial information. Therefore, the analysis and processing of spatial relations in massive data is particularly important. The research of spatial relations has been more than 20 years. Spatial relations have been widely used in spatial reasoning and spatial analysis. Many scholars at home and abroad have made a lot of research results. Most of the existing qualitative spatial relationship models are based on sound algebraic systems, using prior knowledge to model artificial, the existing simple spatial relationship model can not fully express the actual application of the spatial relationship; The existing complex spatial relationship model can only be used to identify the specific complex spatial relationship, and it is not universal. Using the method based on the sound algebraic system to construct the spatial relation model artificially in big data, it is necessary to analyze the prior knowledge of the spatial object, and there are many elements to consider, which leads to the complexity of the formal representation system. It is even difficult to formalize expression, which can not meet the need of spatial expression in big data era. In order to solve the above problems in the existing spatial relational models, we study the spatial relational reasoning methods. In this paper, we propose a method to combine qualitative spatial reasoning with machine learning to generate spatial relational models. The method in this paper can generate spatial relation model without prior knowledge, and it has good generality. The main contents of this paper are as follows: (1) in this paper, the background, significance and development of spatial relations in spatial reasoning are discussed. Then it introduces the research content and the organization structure of this paper. (2) it introduces the basic concept of spatial relation and the related knowledge of machine learning, and describes the MLNB algorithm. (3) through the analysis of the problems existing in the existing spatial relationship model, A method of generating spatial relation model is presented, and the flow of this method is introduced. In this paper, the general feature set of spatial relations is proposed, and the domain decomposition, subregion and extended region partition of spatial objects are introduced in detail. On this basis, a feature classification strategy is proposed to group and combine sub-regions in order to reduce the number of features processed. (4) the method of generating spatial relation model proposed in this paper is applied to multiple data sets. The advantages of this method are analyzed by comparing the results applied to different data sets. The method is applied to the spatial relationship recognition of Chinese text, and good results are obtained. (5) the contents of this paper are summarized, and the prospect of further work is put forward.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP181;TP391.1
[Abstract]:With the development of information technology, the amount of information increases suddenly, and the value density of information decreases. The huge amount of information seems to provide more information to people, but in fact, it makes it more difficult for people to use information. Therefore, a variety of big data technologies emerge as the times require, aiming at the specialized processing of massive data. Through the acquisition, storage, management and analysis of the data, it provides people with the information they really need. As a kind of massive information, spatial information is applied to every aspect of human life. Whether it is the push of news on the Internet, the precise delivery of advertisements in e-commerce, the analysis of information in navigation and traffic conditions, the rescue of governments after disasters, or the location of buildings in the development of cities and enterprises, Can not do without big data support and analysis of geospatial information. Therefore, the analysis and processing of spatial relations in massive data is particularly important. The research of spatial relations has been more than 20 years. Spatial relations have been widely used in spatial reasoning and spatial analysis. Many scholars at home and abroad have made a lot of research results. Most of the existing qualitative spatial relationship models are based on sound algebraic systems, using prior knowledge to model artificial, the existing simple spatial relationship model can not fully express the actual application of the spatial relationship; The existing complex spatial relationship model can only be used to identify the specific complex spatial relationship, and it is not universal. Using the method based on the sound algebraic system to construct the spatial relation model artificially in big data, it is necessary to analyze the prior knowledge of the spatial object, and there are many elements to consider, which leads to the complexity of the formal representation system. It is even difficult to formalize expression, which can not meet the need of spatial expression in big data era. In order to solve the above problems in the existing spatial relational models, we study the spatial relational reasoning methods. In this paper, we propose a method to combine qualitative spatial reasoning with machine learning to generate spatial relational models. The method in this paper can generate spatial relation model without prior knowledge, and it has good generality. The main contents of this paper are as follows: (1) in this paper, the background, significance and development of spatial relations in spatial reasoning are discussed. Then it introduces the research content and the organization structure of this paper. (2) it introduces the basic concept of spatial relation and the related knowledge of machine learning, and describes the MLNB algorithm. (3) through the analysis of the problems existing in the existing spatial relationship model, A method of generating spatial relation model is presented, and the flow of this method is introduced. In this paper, the general feature set of spatial relations is proposed, and the domain decomposition, subregion and extended region partition of spatial objects are introduced in detail. On this basis, a feature classification strategy is proposed to group and combine sub-regions in order to reduce the number of features processed. (4) the method of generating spatial relation model proposed in this paper is applied to multiple data sets. The advantages of this method are analyzed by comparing the results applied to different data sets. The method is applied to the spatial relationship recognition of Chinese text, and good results are obtained. (5) the contents of this paper are summarized, and the prospect of further work is put forward.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP181;TP391.1
【相似文献】
相关期刊论文 前10条
1 杜世宏;空间关系模糊描述及组合推理的理论和方法研究[J];测绘学报;2005年01期
2 杜世宏;秦其明;王桥;;空间关系及其应用[J];地学前缘;2006年03期
3 马林兵;曹小曙;;空间关系的动态性和模糊性描述[J];地理与地理信息科学;2006年06期
4 胡圣武;王宏涛;;空间关系的研究进展[J];测绘科学;2007年01期
5 许s，
本文编号：2151685
本文链接：https://www.wllwen.com/jingjilunwen/dianzishangwulunwen/2151685.html
