【作者】 江宝得；

【导师】 吴信才；

【作者基本信息】 中国地质大学 ， 地图制图学与地理信息工程， 2013， 博士

【摘要】 土地利用数据库是对土地利用现状进行详细调查所获得的专题型数字化成果数据库。主要用于土地资源管理,以反映土地的地理位置、数量、质量、空间分布、土地利用类型、权属和利用情况等。与一般的地理数据类似,土地利用数据包含空间数据和属性数据两部分,其中,土地利用空间数据中的地类图斑具有覆盖整个空间区域无缝隙的特点,土地利用属性数据蕴含丰富的语义信息和多层次类型划分的专题特征。因此,土地利用数据库制图综合有别于传统的普通地图数据库制图综合,它是制图综合理论与技术在土地利用专题数据领域的典型性应用,其目的是为了提高地图的可读性和可理解性。在制图综合处理过程中,除了要考虑传统的制图理论因素外,还需要根据应用目标需求和制图区域的特点,综合考虑该区域地类图斑分布模式、统计特征、语义、拓扑、图形、精度等多种约束因素,按照综合的先后约束顺序进行系统化的综合处理。目前,很多学者对土地利用数据制图综合进行了大量的研究,取得了一系列研究成果。但是,这些研究在综合方式上主要还是以交互式综合为主,智能化程度不高；在综合算法上,由于实施多重约束的困难性,算法通常都只顾及了某种约束条件或侧重于综合的某一方面,导致顾此失彼,而土地利用数据制图综合是一个系统工程,需要从多个方面综合考虑各种约束因素。因此,分析研究土地利用数据制图综合的多重约束因素,探讨顾及多重约束条件的面向土地利用数据库制图综合的自动综合算法和研究智能化的综合系统在理论和实际应用中都具有非常重要的意义。本文主要以土地利用数据库中的地类图斑为研究对象,考察其在综合过程中应遵循的合理综合流程、被实施的主要综合操作以及各种操作需要顾及的多重约束条件；按照综合操作实施的先后顺序,选取图斑归并综合、狭长图斑降维综合和图斑化简综合三种最常用的图斑综合操作,研究其在多重约束条件控制下的自动综合方法,并构建智能化的土地利用数据库综合系统,通过1：10000到1：50000的土地利用数据自动综合实验和对实验结果进行评价分析,验证本文提出的自动综合算法的有效性和合理性以及智能化综合原型系统的可行性和实用性。具体研究工作如下：1)介绍了本文的研究目的和意义,总结了普通地图制图综合的研究现状和土地利用图制图综合的研究现状,指出了当前面向土地利用数据制图综合研究的不足,界定了本文的研究范围及研究内容。2)介绍了制图综合约束的基本概念和制约制图综合的因素,分析了土地利用数据的基本特征：多层的空间数据组织结构、地类图斑全覆盖和丰富的语义信息和专题特征；浅析了土地利用图与普通地图制图综合的区别；从尺度约束、结构约束、空间关系约束、语义约束、几何约束、综合过程约束六个方面构建了面向土地利用数据库制图综合的约束控制知识体系,并分别阐述了各个约束的含义和意义;从综合质量控制和综合的完整性角度考虑,给出了几种针对土地利用数据综合结果进行整体评价的方法,分别是：面积变化率、语义相似度、结构相似性和布局稳定性,以便检查综合结果是否合理,从而保证综合质量。3)针对地类图斑自动归并综合,分析了图斑归并综合需要顾及的多重约束条件以及各约束条件实施的先后顺序,并根据图斑间的拓扑关系进一步将图斑归并综合划分为聚合、合并和融合三种类型,分析了每种类型的含义及操作方式,并指出融合操作可视为合并操作的一种特殊情况,从而将整个归并综合分为聚合和合并两大操作步骤,并分析指出聚合操作应先于合并操作实施；接着针对图斑聚合综合,提出了一种基于栅格与矢量数据模型的图斑自动聚合综合方法,该方法的思想是：先基于栅格数据模型对待聚合图斑进行自动空间区域划分及聚类分析,然后基于矢量数据模型对图斑实施自动聚合综合,从而很好的顾及了图斑的空间分布约束、语义约束、图形结构约束、精度约束等多种约束条件；针对图斑合并综合,给出了一种基于图斑临近关系分析的合并方法,其中,临近关系的度量是关键,介绍了一种顾及拓扑邻近和语义邻近的度量方法,并在此基础上,介绍了基于邻近度量的图斑合并方法以及各种地类合并应遵循的优先顺序;最后,给出了图斑归并综合的一体化自动综合流程与实验结果。4)针对狭长多边形图斑自动降维综合,分析了狭长多边形图斑需要作降维综合处理的原因以及降维综合产生的不一致性问题,指出了现有研究的不足;介绍了一种基于CDT的狭长多边形图斑自动探测方法和图斑局部狭长部分自动分割方法,在此基础上,给出了狭长多边形图斑自动降维综合的方法；接着针对狭长多边形图斑降维综合后导致的拓扑不一致性和语义不一致性问题,分别给出了一致性自动改正方法,前者在狭长多边形图斑与其邻近多边形图斑之间的拓扑关系的基础上,将弧段分为三类,分别用Ⅰ、Ⅱ、Ⅲ表示,并指出狭长多边形图斑降维综合后的拓扑一致性改正就是保持Ⅰ类弧段不变,Ⅱ类弧段删除,Ⅲ类弧段延长到中心线并相交打断后局部拓扑重建的处理过程；后者在土地利用数据语义综合规则的基础上,对狭长多边形图斑的邻近多边形进行语义一致性改正；最后,给出了狭长多边形图斑自动降维综合的实验流程与实验结果。5)针对图斑自动化简综合,在详细分析各种约束因素对图斑化简综合产生的影响基础上,将图斑按照所表达的地理要素特征分为自然地物和人造地物两大类,并分析了两类图斑各自的特点以及化简的优先顺序；接着针对人造地物图斑化简综合,在现有基于最小二乘法平差原理的建筑物图斑化简算法基础上,提出了一种适用于土地利用人造地物图斑化简综合且顾及多重约束因素的改进型图斑化简综合算法,该算法先对人造地物的形状结构进行探测,然后对以共享边存在的人造地物图斑进行化简前预处理,最后根据化简综合所需顾及的多重约束条件建立多个约束方程,联立方程组进行最小二乘平差解算得到最终化简结果。针对自然地类图斑化简,指出了拓扑关系、图斑边界弯曲特征和面积平衡保持是自然地类图斑化简时需要重点考虑的约束因素,并根据图斑是否被其它图斑包含,又将自然地类图斑分为岛状图斑和非岛状图斑两种,对于非岛状图斑,提出了一种基于图斑边界弯曲特征分析并顾及多重约束条件的自然地类图斑自动化简方法,该方法先利用CDT对自然地类图斑的共享边界进行弯曲探测和识别,分析其凹凸特性,然后运用渐进式化简综合思想,对共享边界进行凹凸交替式删除化简,使得在化简过程中图斑的面积尽可能的保持了平衡,最后利用夸大或删除部分弯曲来平衡化简积累的面积差,从而得到面积保持平衡的化简结果;对于岛状图斑,先用最小外包矩形对图斑边界实施分割,并对每一段按照非岛状图斑的化简策略实施化简,从而在顾及面积与弯曲特征的约束基础上,很好的顾及了图斑主方向及整体形态约束特征。为了检验图斑自动化简算法的有效性,最后给出了图斑自动化简综合的实验流程与实验结果。6)针对土地利用数据库制图综合的实际应用需求,进行了原型系统分析与设计,并在MapGIS K9平台上用MS Visual Studio2005进行二次开发,实现了满足土地利用数据库制图综合的原型系统Landuse-Gen:该系统采用工作流的设计思想,将综合过程中所需的源数据、综合算子、综合参量、综合知识、综合规则以及其它所需资源视为综合流程中的—个节点,存放在系统的功能仓库中,制图人员根据项目预先制定的综合操作方案,将综合过程中所涉及的流程节点通过系统提供的流程搭建工具搭建成综合流程,每个流程针对—个特定的综合应用需求,其构建的过程即是制图专家综合决策的体现,因此,每个综合流程在执行时,即被赋予了人的部分知识、经验和决策能力,从而使得系统具备了一定的智能性。为了检验本综合系统的可行性和实用性,详细介绍了以广西铁山港区1：1万土地利用数据为例进行的1：5万的综合试验情况。更多还原

【Abstract】 Land-use database is a thematic type of digital map databases which is based on a detailed survey of land-use status. It is mainly used for land resource management to reflect the land location, quantity, quality, spatial distribution, land-use type, ownership and use status etc. The same as the general geographic data, land-use data contains spatial data and attribute data, the polygons in the land-use spatial data have the characteristic of covering the entire space region with seamless distribution, and the land-use attribute data are enriched with semantic information and multi-level thematic characteristics. Therefore, land-use database cartographic generalization is different from common map database cartographic generalization, it is the typical application for cartographic generalization theory and technology taking into the field of land-use thematic data, its purpose is to improve the map’s readability and understandability. In the process of cartographic generalization, it should consider the regional distribution patterns of sporadic parcels, statistical characteristics, semantics, topology, graphics, precision constraints and other factors according to the application target requirements and mapping regional characteristics, in addition to considering the traditional cartographic theory factors, it needs to generalize systematically in a comprehensively constrained order. Many scholars have done a lot of research on land use data generalization, and made a series of research results, but these studies were mainly, on the one hand, on an interactive generalization for operation lacking of intelligence, and on the other hand, these algorithm usually only took into account a certain constraint or focused on a particular aspect of generalization, due to the difficulty of implementing multiple constraints, which made the result attend to one thing and lose another, and the land-use data generalization is a systematic project, it needs to consider the various constraints. Therefore, neither in theory nor in practical applications, it is meaningful to study on multiple constraints for land use data generalization, and make research on new methods and intelligent system.This thesis focuses on sporadic parcels of land-use database, examines its rational procedure that should be kept to in the generalization flow, the operation that generalization should be taken, and the constraints that operation should be taken into account. According to the operations’ order, polygons’merging generalization, narrow and long polygons’dimension-reducing generalization and polygons’simplifying generalization which are the most common operations for the polygons’ generalization are taken into account, studying on the automatically generalizing methods under the control of multiple constraints and building an intelligent land-use database generalizing system, a land-use data generalization experiment is executed from scale1:10000to1:50000, and the generalization results are assessed which is to verify the proposed algorithm whether is effective and rational and the intelligent generalizing system whether is feasible and practical. Specific research works are as follows:1) The purpose and significance of this research is introduced, then, the general map cartographic generalization research status quo and land-use map generalization research status quo are summarized, the lack of current research on land-use data generalization is pointed out, and the scope of this thesis and the research content are defined.2) The concept of constraint for cartographic generalization and the impact of constraint on cartographic generalization are introduced. The basic characteristics of land use data that multilayer spatial data organization, to cover the whole region by sporadic parcels and rich semantic information and thematic information are analyzed. The difference between land-use map cartography generalization with common map cartographic generalization is introduced. The systematic knowledge of constraints for land-use database cartographic general ization consist of six aspects, such as scale constraint, structural constraint, spatial relations constraint, semantic constraint, geometric constraint and operation process constraint, and each constraint’s meaning and significance is elaborated. Considering the quality controlling and integrity of generalization, several methods for land-use database generalization result assessment are proposed, including area change rate, semantic similarity, structural similarity and distribution stability, which is in order to check whether the results are reasonable, and to ensure generalization quality.3) As for sporadic parcels automatically merging generalization, what constraints that should be taken into account as well as the order that constraints should take for sporadic parcels merging generalization is analysed, and the merging generalization is divided into three types according to the topological relations between polygons, their are aggregation, consolidation, and amalgamation, each type’s meaning and method of operation is given, it points out that the amalgamation operation can be considered a special case of the consolidation operation, and that the aggregation operation should be taken before the consolidation operation. Then, a method for sporadic parcels aggregation generalization is proposed based on raster and vector data model, firstly, it divides the sporadic parcels into several spatial zones and takes cluster analysis base on raster data model, secondly, it aggregates the cluster polygons based on vector data model. The method takes into account the spatial distribution constraint of polygons, semantic constraint, graph structure constraint, precision constraint and other constraints. Another method for sporadic parcels consolidation generalization is proposed based on neighborhood relationship analysis of parcels, in which the measure of neighborhood relation is the key point. A method for measuring the degree of neighborhood considering the topology and semantic constraints is introduced, and on this basis, the order for the terra-type parcels consolidation to take is given. Finally, the experiment flow for sporadic parcels automatically merging generalization and results are presented.4) As for long and narrow polygon automatic dimension-reducing generalization, the reason that why do the long and narrow polygons need to take dimension-reducing generalization and what inconsistencies would be made by the dimension-reducing generalization is given, the lack of current research is pointed out. An method for detecting long and narrow polygon automatically and an method for segmenting the long and narrow part of polygon automatically witch are all based on CDT is proposed, on this basis, a method for long and narrow polygon automatic dimension-reducing generalization is introduced. Then, focusing on the inconsistency of topology and semantics produced by the dimension-reducing generalization, an approach for the inconsistencies correcting is given separately. The previous is based on the topology between the target polygon and its adjacent polygons, it classifies the arcs into three types, Ⅰ, Ⅱ and Ⅲ, and presents that the topology consistency correcting is keeping the type I arcs, deleting type II arcs, and prolong the type III arcs to the polygon’s centerline, then, clips the arcs and rebuild the topology partly, and the latter aims at semantics consistency correcting based on the sematic rules. Finally, the experiment flow for the long and narrow polygon automatic dimension-reducing generalization and results are presented.5) As for sporadic parcels automatically simplifying generalization, based on the detailed analysis of various constrained factors that impact on the polygons simplification, it divides the sporadic parcels into two major categories that are natural features and man-made features according to the geographic features of polygons’ expression, and the characteristics of each type of polygons and simplification priorities are analyzed. In the following, an improved version of least squares adjustment for man-made features simplification algorithm considering multiple constraints is proposed base on the principle of building polygons simplification algorithm, the algorithm first detecting the structure of man-made feature, then pretreating the polygon before taking simplification that whose edges are shared by another polygon, the last, constructing various equations according to the simplification required multiple constraints, and solving the equations under the least-squares adjustment to get the final results. For natural features simplification, it points out that topological relation, the curved character of polygon’s boundary and area balance maintain are mainly constraints that natural features simplification should take into account, and it divides the polygons into island polygons and non-island polygons according to whether the polygon is contained by other polygon. For the non-island polygons, an automatically simplifying algorithm is presented base on the curved character of polygon’s boundary and considering multiple constraints, it first detecting and identifying the bend character of polygons shared borders base on CDT, and analyzing its convex characteristics, then taking a progressive simplification policy, it deletes the convex bend alternatively to keep the polygon’s area balance as possible as it can in the process of simplification, and finally exaggerating or deleting some bend to balance polygon’s accumulation area of simplification, getting the result of simplification with polygon’s area kept balanced. For the island polygons, first splitting the polygon’s boundary with its smallest outsourcing rectangle, then taking the policy of non-island polygon’s simplification for each section, thus keeping the polygon’s mainly direction and overall structure characteristics well in addition, expect containing the base area balance and bending characteristics constraints. In order to test the effectiveness of automatically simplifying algorithm for polygons, an experimental flow of polygons automatically simplifying generalization and results are presented at last.6) A prototype system is analyzed and designed based on the actual application requirements for land use database cartographic generalization, and the prototype system called Land use-Gen which meet the land use database cartography is developed on the MapGIS K9platform using MS Visual Studio2005. The system takes a workflow policy design, which takes the required source data, generalization operators, generalization parameters, generalization knowledge, generalization rules and other necessary resources as a flow node in the generalizing process and each node is stored in the system’s functional warehouse. The cartographers, according to the pre-established solution of generalizing operation for project, build the generalizing work flow with the referring flow node in the process of generalization using system’s providing build tool, and each work flow is for a specific generalizing application requirements, it is a embodiment of cartographic experts decision-making for generalization. Therefore, when each work flow is executed, the generalizing process is endowed with some knowledge, experience and decision-making ability of people, and thus making the system have a certain intelligence. In order to check the system’s feasibility and practicality, an experiment conducted with Tie shan gang land use data of Guang xi for generalization from1:10000to1:50000is described in detail.更多还原