【作者】 陈长林；

【导师】 翟京生；

【作者基本信息】 解放军信息工程大学 ， 地图制图学与地理信息工程， 2013， 博士

【摘要】 数字化解决了数据从无到有的问题，而信息化则需要进一步实现数据的融合共享，以达到数据的一致性、实时性、高效性。我国的海图生产虽然已经实现了数字化，但是相比于发达国家的海图生产水平而言，存在着产品数据不一致、更新周期相对较慢和生产效率难以较大提高等问题。针对这些问题，本文提出了一个新的海图生产体系，并对其关键技术进行了研究。论文主要工作和取得的成果有：1．对现代海图生产过程中涉及到的“海图生产体系”、“海图数据模型”、“自动综合算法”和“空间数据匹配”等四个关键问题的国内外研究现状进行了评述和分析。2．提出了同源多尺度海图生产体系的概念、框架结构，分析了其主要生产工序和关键技术。该生产体系通过多源海图数据的融合，保证源数据的唯一性、现势性和产品数据的同宗同源；同时允许产品数据之间存在派生关系，避免了因源数据和产品数据的比例尺跨度过大导致数据生产的困难；引入了“空间几何实例共享池”，实现不同产品数据之间空间几何数据的共享，保证了产品数据之间的一致性。3．针对实体模型存在数据冗余、容易导致数据不一致，而拓扑模型则过于复杂以及存在关系冗余的问题，本文提出了一个无缝无级空间数据融合模型，不仅能够实现多源空间数据的融合，而且避免了传统空间数据模型中的数据冗余和不一致等问题；结合S-100和S-57两个海图数据国际标准，给出了海图要素模型和海图属性模型，提出了“最小拓扑”空间模式；同时，依据海图源数据融合模型，提出了一套海图源数据库设计与实现的方案：通过海图概念字典和要素目录实现海图数据语义的统一，通过要素集合和空间集合的解耦合实现了海图源数据库的无缝无级表达。4．针对外部数据与海图源数据不一致的问题，本文提出了一种基于XML，由代词、名词、形容词、条件语句、操作语句等语法结构构成的转换语言，实现了要素和属性的自动转换，并通过拓扑重构技术实现了空间模式的自动转换，进而实现了DNC和CDC数据等外部数据向海图源数据的转换；研究了新数据与源数据库中要素的自动匹配方法，给出了一种“基于误差理论的改进型Hausdorff距离”；最后分析总结了不同海图源数据之间的质量评价因子和质量评价规则知识。5．针对从源数据库中派生产品数据的情况，本文重点研究了海图水深的自动综合方法。通过对现有水深点群综合方法的分析，发现传统方法将水深点群视为平面点集、过分强调空间分布特征，导致水深自动综合难以实用化；针对现有水深点群自动综合方法的不足，提出了“将水深点群视为空面点集、以海底地形特征识别为基础进行背景水深自动综合”的新思路；通过建立Delaunay三角网和局部动态更新技术，以“离散曲率最小”、“平均间距约束”、“深度变化约束”及其组合方式为约束条件，实现了基于TIN空间曲面的水深点群渐进式自动综合。6．针对海图源数据库出现更新时，需要将变化信息传递至产品数据的情况，本文重点分析了产品数据中曲线的更新方法。利用曲线约束Delaunay三角网建立CDT-Btree，并通过伪弯曲剔除、重采样等优化以及并列弯曲识别和调整等方法，构造了一种与视觉感受相一致的曲线弯曲特征CDT-Mtree；以CDT-Mtree为基础，通过多级弯曲匹配，实现了源曲线和产品曲线的弯曲匹配和变化识别，进而实现了产品曲线中变化弯曲的增量更新。7．给出了一个同源多尺度海图生产的实验。以DNC和CDC实验数据为基础构建了海图源数据库原型；对ENC产品规范中关于数据内容的规定进行了总结；以ENC为例，给出了从海图源数据库中生产ENC新数据和更新数据的基本示例，初步验证了同源多尺度生产的可行性。更多还原

【Abstract】 Digitalization means converting paper data into digital data, while informatization aims tointegrate data based on digitalization, to ensure consistency, updating in time and highefficiency of the data. China charts production has been realized digitalization, howevercompared with developed countries, some technologies are still lagging behind, such as lowefficiency, not timely data update as well as data inconsistency in different products. In order tosolve these problems, a solution is proposed in this dissertation, as well as its key problems.The main works are as following:1. Four issues related to modern chart production－"chart product architecture","chartdata model","auto generalization algorithm","matching algorithm of spatial data" aresummarized and analyzed in the front of this dissertation.2."One Source, Multi Representation" Chart Production System is proposed, including itsconcept, theory architecture, main process and key technique.By multi-source chart data fusion,this production system can ensure all the source data be unique, update in real-time, all theproduct data be consistent; it allows product data derivating from another product data, whichavoid difficulty from producing product data from source data of larger scale;"Geometry SharePool" realizes geometry share among different product data, which ensure all the product databe consistent.3."Entity Model" is redundant in geometry and has problem of topological inconsistency，and "Topology Model" is redundant in topological relation and its realization and operation isvery difficult. So, an seamless, scaleless, not layered, and irredundant geospatial data model－"Seamless, Scaleless Source Data Fusion Model" is promoted in this dissertation, which can notonly integrate various chart data,but also avoid data redundancy or relation redundancy intraditional data models. Considering the international chart standard of S-57and S-100, afeature data model and attribute data model of chart is given, and a spatial data model withminimized topology is brought out too. According to "Seamless, Scaleless Source Data FusionModel", a schema of chart source database is put forward: chart data semantic is unified byestablishing chart concept collection and features catalogue, seamless and scaleless chart sourcedata is constructed by decoupling feature collection and spatial geometry collection.4. After analyzing the transfer relationships of different chart data, the dissertation putsforward a XML-based transfer language, which is consisted of pronoun, noun, adjective,condition sentence, operation sentence and other syntax structures, feature and attribute can betransferred automaticlly by this language; by topology reconstruction, spatial schema can be transferred automatically. With the XML-based transfer language and topology reconstruction,transforming from CDC, DNC and other exterior data to chart source data can be the realized.Automatic feature matching method between new data and data in source database is alsoresearched, and a new line feature matching method based on "improved Hausdorff distancebased on error theory" is put forward in this dissertation. At last, chart data quality evaluationcriteria are given.5. In order to produce product data from source database, chart soundings’ selectionmethod is researched in this dissertation. By analyzing existing research on soundings’ selection,we can draw a conclusion that the traditional soundings’ selection method taking soundings asplane points and lay stress on spatial distribution characteristics, which causes soundings’ autoselection difficult to practical; and space points’ selection, and then an idea of how to selectsoundings is put forward. In view of existing shortcomings of soundings’ auto selectionmethods, an idea of select background soundings in space points, in order to preserve seabedcharacteristic, is advanced, then based on Delaunay triangle net and local dynamic update, aprogressive method of soundings’ auto selection base on TIN surface is put forward, thenseveral experiments with "minimum curvature limit","average spacing limit","depth valuechange limit", and all these limit combined are carried out and compared, from which we candraw a conclusion that sounding selection in combined limit has better effect.6. In order to update product data once source database changes, a method to update curvein product is researched. Based on construction of Constraint Delaunay Triangle Net, Bend-Binary-Tree of curve can be built, that is called CDT-Btree. Optimization of CDT-Btree can becarried out by resembling, branch node adjustment, then a Multi-way Tree, called CDT-Mtree,can be built after neighbor bends recognized, which represents the curve bend characteristiccompatible with visual perception. Based on CDT-Mtree method, an incremental updating andincremental generalization method is proposed with the basic element of curve bend. With themulti-level curve bend matched, the change between source curve and product curve bend ismatched and recognized.7. An experience conforming to "One Source, Multi Representation" Chart Production iscarried out. A basic chart source database is built from DNC and CDC experiment data; themain regulation on data content of ENC product specification is introduced; then twoexperiment about chart production and update from source database is introduced taking ENCas an example, which proved the feasibility of "One Source, Multi Representation" chartproduction.更多还原