Spatial Data Organization And Indexing Research Driven By Spatial Distribution Pattern

Spatial data organization and management is the foundation of spatial information technologies applications. Spatial data appears an explosive increasing trend because of the rapid development of smart city, VGI and LBS applications. Currently, spatial big data have various types, large volume, and quick incrementation and are characterized by multi-dimension domains and always unstructured as well. Under these circumstances, existing data organization and index methods have some problems in different applications such as low efficiency in query and analysis, high cost of data storage and transmission. Building reasonable and efficient methods in data organization and management is pivotal to push forward spatial data application and analysis.Spatial database is the main place of data management. Related research topics in this area focus on spatial data model, spatial data organizing, index and operations .etc. However,there are three main key issues in spatial database research: ?Space partition, neither space-driven "object-ignoring space partition" nor object-driven "space-ignoring object division"cannot fully considered group positioning of geographical objects. This limits the adaptability of space partition; ?Data organizing and storage. Current organizing methods are based on discrete storage of geographical objects. Relationships among geographical features are isolated. These methods ignore spatial proximity and spatial heterogeneity and cannot solve problems between multi-dimensional spatial data and one-dimensional structure of computer. This will result in low efficiency of data organization. And data organizing lacks abilities in geo-computation supporting,geographical knowledge finding and geographical scenes simulation;?Spatial index and operations. Existing balanced tree index structures get high query performance in the cost of big constructing and updating.Unbalanced tree structures have high efficiency in constructing and updating. But it cannot solve problems in data skewing with aggregated distribution pattern. And it ignores importance of spatial data organizing and storage model. Lacking unified efficient spatial operations algorithms in spatial index and organizing. In a word, current spatial data organizing methods focus on computer implementation. They pay little attention to geographical disciplines and geoinformatic knowledge. The problems between structured linear storage of computer and unstructured multi-dimensional spatial data still exist. These methods cannot meet efficient massive spatial data organizing and management in big data era.For breaking the limit of current spatial data organizing, storage and index which is characterized by object discretization and structured storage, this paper carries out research in above three key issues.The main content and results of this paper can be summarized as following:(1) Spatial distribution detection and space partition methods. Starting from spatial proximity and heterogeneity, this paper introduces index in area of divided sub-district and perimeter variation coefficient based on spatial distribution classification and quantitative parameter description. Then fast spatial distribution detection method is proposed. After that,quantitative parameter description local density p and exclusion group distance ? for spatial heterogeneity respectively is led into this research. And local clustering space division method is designed for multi-target optimization with aggregated spatial distribution pattern.Also, outlier object optimization based on decision diagram is implemented in this paper.(2) Hierarchical nested organizing model designing considering spatial proximity and heterogeneity. Starting from impact mechanism and constraint rules of spatial proximity and heterogeneity on spatial data organizing, this paper designs hierarchical nested organizational model based on document-oriented non-relational model against geographical complex entities with hierarchical nested structure and interaction of diverse geographical phenomenon. These methods upward achieve nested packaged organizing among models, and downwards implement free patterns and easy expansion of unstructured organizing inside the model.(3) PatternList spatial index. This paper analyzes influence of data structure in spatial index query, update and maintenance. And use jump table structure to construct semi-equilibrium spatial index of PatternList. Besides, query, insert and delete operations flow and analysis on time complexity of them is proposed. Finally, combined with hierarchical nested model spatial data operation algorithm and dynamic batch updating under the support of PatternList is implemented.These three main research contents are closely related and progressive. Spatial distribution detection and space partition is the foundation of whole research. Hierarchical nested organizing model is organizing and storage mode of spatial data. Spatial index structure PatternList is the mapping of spatial partition and hierarchical nested organizing model. Ultimately, VGEs prototype System is constructed to implement hierarchical nested organizing methods, semi-balanced spatial index structure of PatternList and related spatial operations algorithms. Besides, typical testing data is used to verify effectiveness and efficiency of space partition method, hierarchical nested organizing model and spatial index structure that have been proposed in this paper.This study focuses on improving crucial technologies in spatial data organizing and storage. It provides sustentation of ideology and methodology for researches in spatial data organizing and index. Research productions of this study not only enrich methods for spatial index and organizing, but also improve spatial data service ability of GIS.