空间联机分析模型驱动的时空数据探索方法研究

Aiming at EB-level and multidimensional spatio-temporal data volume generated by the era of Space-Air-Ground integration stereoscopic observation in China,how to discover the knowledge and pattern of data from different time and space dimensions rapidly and effectively becomes a problem.And this problem has grown up to be a huge challenge to GIS spatial analysis. Concerned with the limitations of continuous geographic spatio-temporal data expression and efficient data analysis provided by the existing Spatial Online Analytical Processing model, this project will construct the multidimensional spatial on-line analytical computing paradigm. Then proposes a mapping mechanism constructed from Spatial Online Analytical serial algorithm to Distributed Parallel Processing Architecture on the basis of it.The main research includes the following sessions: designing an abstract data model for Spatial OnLine Analytical Processing by considering different spatial distributions and defining multidimensional spatio-temporal cube algebra and its operation primitives,and also designing a low-dimensional and high-density spatio-temporal data organization method for distributed computing,then realizing a mapping mechanism constructed from spatial online analytical serial algorithm to distributed parallel processing architecture,and then instantiating and validating the model on the basis of exploratory spatio-temporal data analysis scenarios including spatio-temporal characteristics of urban atmospheric pollutant concentration,finally summarizing the general process of spatio-temporal data exploratory analysis which applies the spatial online analysis model.Its aim is to provide a new solution to the optimization research on exploratory analysis of multidimensional spatio-temporal big data.

面对我国空天地一体化立体观测时代所带来的EB级多维时空数据量，如何及时有效地从不同时间与空间维度挖掘出数据所蕴含的知识与模式，已经成为GIS空间分析方法面临的巨大挑战。针对现有空间联机分析模型在支撑连续型地理时空数据表达和高效数据分析时存在的局限，本项目在构建多维空间联机分析计算范式的基础上，系统研究空间联机分析串行算法到分布式并行处理架构的映射机制，主要研究内容包括：建立面向连续型空间分布形态的多维空间联机分析抽象数据模型，定义多维时空立方体代数及操作原语，设计面向分布式计算的低维高密度时空数据组织方法，实现多维空间联机分析算子到分布式异构计算框架的映射机制，并基于城市大气污染物浓度时空变化特征的探索性时空数据分析场景进行模型实例化表达和验证，归纳出应用空间联机分析模型进行时空数据探索性分析的通用流程，旨在为多维时空大数据探索性分析方法的优化研究提供一种新的解决思路。

面对我国空天地一体化立体观测时代所带来的EB级多维时空数据量，如何及时有效地从不同时间与空间维度挖掘出数据所蕴含的知识与模式，已经成为GIS空间分析方法面临的巨大挑战。本项目研究针对现有空间联机分析处理模型在支撑连续型时空数据表达和高效数据分析时存在的局限，构建时空维分离的六元组立方体数据模型，定义支撑基础、拓展和连续型三类原语操作的多维空间联机分析算子，设计空间联机分析中时空立方体的物化方法、数据立方体计算的优化方法和查询加速策略，实现支持空间联机分析操作原语的时空地图代数算子到分布式异构计算框架的映射机制。并以“城市大气污染物浓度时空变化特征”的探索性分析场景为例，构建空间联机分析高性能计算原型系统进行模型实例化表达和验证，归纳出应用空间联机分析模型进行时空数据探索性分析的通用流程，为多维时空大数据探索性分析方法的优化研究提供一种新的解决思路。