节点文献
空间信息复合分析模型研究
Study on the Models for Integrated Analysis of Spatial Information
【作者】 薛丰昌；
【作者基本信息】 中国矿业大学 ， 地图制图学与地理信息工程， 2008， 博士
【摘要】 现有GIS空间分析技术不能满足多维信息空间分析的需求,如何将空间依赖、空间尺度效应、空间非均质性共同结合到空间分析中缺少有效模式,空间分析中数据的多维复杂性往往被忽略。空间信息在信息相互作用中存在着尺度效应、空间依赖性、空间异质性,多源多维空间信息耦合关系表现为空间依赖性和空间非均质性作用下多尺度空间要素的信息与知识的跨尺度转换、推绎的关系,空间信息复合分析的实质是应用多维、多源信息,采取多种分析方法对事物进行综合性认识的过程。将泛布尔函数引入空间信息复合分析中,建立了空间信息泛布尔函数复合分析模型。提出了空间属性多态化、空间效用函数、泛布尔空间逻辑关系的知识化表达等技术理念,建立了基于空间信息泛布尔函数复合分析模型的空间反演模型与空间预测模型。建立了空间信息分层线性复合分析模型。地理学第一定律的提出对解决空间依赖性起到了良好的指引作用,形成了地理加权回归模型(Geographical Weighted Regression,GWR),GWR模型是未考虑尺度差异的线性组合模型,因而存在着信息损失。空间信息分层线性复合分析模型从空间数据组内效应、组间效应分析空间过程,通过进行“回归的回归”两阶段分析综合反映空间依赖关系和空间尺度效应对空间过程的影响。建立了空间信息多维尺度复合分析模型。现实世界中一组空间观测信息构成的信息模式往往受到一些关键性的隐含因素的影响,这些关键因素构成了一个隐含结构,这个结构能够以简化方式但又有效地对信息模式进行表达。多维尺度空间信息复合分析模型将数据变换到相似性空间,根据空间相似性进行分析,在不依赖诸多先验假设情况下,发现并学习数据集的内在规律与性质。以徐州市征地区片划分所涉及的空间要素复合分析为应用实例,对本文建立的空间信息复合分析模型进行了比较分析。结果表明:空间信息泛布尔函数复合分析模型易于实现,能够融合主、客观分析方法,适合应用于对资源环境进行初步评价;空间信息分层线性复合分析模型计算过程紧紧围绕空间依赖和空间尺度效应关系的建模表达,能够较好地揭示空间信息的复合机理;空间信息多维尺度复合分析模型是一种客观的、完全以数据为驱动的探索多维、多尺度空间数据潜在流形结构的过程,适宜于发现原始数据集所蕴含的内在规律与性质。论文建立的空间信息复合分析模型能够从不同视角、针对空间数据的不同特征建立分析评价模式,更加综合、全面利用空间数据的有效信息,有助于解决空间分析中数据的多维复杂性涉及的尺度效应、空间依赖性和空间异质性综合作用问题。
【Abstract】 Spatial analysis of GIS can not fulfil analysis involving multidimensional spatial information. There is no effective model to integrate spatial dependence, spatial heterogeneity and spatial scale effect. Multidimensional complexity is ignored in most spatial analysis.Due to spatial dependence, spatial heterogeneity and spatial scale effect, coupling relations of multidimensional spatial information represent as transition and deduction of multi-scale spatial factors between spatial scales affected by spatial dependence and spatial heterogeneity.The model for integrated analysis of spatial information (IASI)based on pan-boolean fuction is constructed,which includes dispersing spatial attribute, constructing spatial avail function and expressing knowledge of spatial pan-boolean logic relations. Based on these, spatial inversion transformation model and spatial forecast model based on pan-boolean logic are constructed.Multi-level integrated analysis model of spatial information is constructed.The first principle of geography makes a way to solve spatial dependence and geographical weighted regression (WGR) is brought forward.GWR is a linear combination modle not considering spatial scale difference,therefore information losing exists in data processing.In multi-level integrated analysis model of spatial information, spatial process is taken as the process affected by inner and external effect of data and relations of spatial dependence and the spatial scales are distinguished by two regressions.Multidimensional scaling integrated analysis model of spatial information is constructed.The information model deducing from the observation of the real world always affected by some pivotal crytic factors .Those factors compose a crytic structure which can express information model effectively in simplified way.Spatial data is translated to a similar space in multidimensional scaling integrated analysis model of spatial information,the character and regulation of the spatial data can be discovered by analysing spatial similarity,which not depending on much transcendent hypothesis.Taken partitioning land requisition blocks of xuzhou city as an example, three integrated analysis models of spatial information above are compared.It turns out that i.the model for integrated analysis of spatial information based on pan-boolean fuction is prone to realize and can integrate subjective and objective analysis,which fits for elementary evaluation of environment and resource. ii. multi-level integrated analysis model of spatial information emphasizes setting up model to express spatial relations of spatial dependence and spatial scale effection,which fits for discovering integrating principle of spatial information. iii. multidimensional scaling integrated analysis model of spatial information is an objective model operated by data ,which emphasizes finding the cryptic manifold of spatial data and fits for discovering the character and regulation hiding in spatial data.