返回
类型 应用研究 预答辩日期 2018-03-07
开始(开题)日期 2013-01-22 论文结束日期 2017-11-13
地点 东南大学四牌楼校区李文正北楼106会议室 论文选题来源 中央、国家各部门项目     论文字数 8.2 (万字)
题目 嵌入式GIS空间数据多尺度表达与快速显示技术研究
主题词 嵌入式GIS,土地利用现状数据,遥感影像,多尺度表达,LOD
摘要 以智能移动终端为载体的地理信息应用已成为发展趋势,并成为当今地理信息科学领域的研究热点之一。随着地理信息获取手段的日益增多,如智能机、无人机、遥感卫星等,海量空间数据以GB级速度不断呈现。当前计算机硬件制造业的迅猛发展使得移动终端(嵌入式设备)性能在计算速度、内存大小和存储容量等方面都得到较大的提升。但是与空间数据的数据量相比,其计算资源和存储资源依旧是相对有限的,仍然无法满足大数据量空间数据快速处理的需求。嵌入式设备性能的提高速度难以匹敌空间数据的增长速度,这一矛盾也使得地理信息服务质量受到制约。如何利用嵌入式设备有限的计算资源和存储资源进行海量空间数据存储、处理与显示是一个意义重大且迫在眉睫的问题。 空间数据多尺度表达可以实现不同细节层次信息的提取,已成为嵌入式设备上空间数据快速处理与显示的一种有效方法。我国国家自然科学基金委员会在构建“数字中国”地理空间基础框架的总体战略中将空间数据的多尺度表示作为地理空间数据库关键技术研究的核心问题之一。现有研究虽然已取得不少成果,但仍存在一些亟待解决的问题:在矢量多尺度信息提取过程中没有很好地融入专题数据知识规则,仅考虑了数据的空间结构特征;基于商业数据库或分布式文件系统的数据存储管理方式对硬件配置要求高且管理复杂,不适用于嵌入式设备中影像数据的高效存储与管理。 本论文致力于研究面向嵌入式的空间数据多尺度表达与快速显示关键技术,突破了嵌入式设备上百兆级矢量数据以及GB级遥感影像数据快速显示技术难题;研发了一套数字化动态变更外业调查与核查系统,解决了土地外业实地调查中定位、底图和属性信息的一体化集成技术问题。论文主要研究成果包括以下五个方面内容: (1)基于特征的矢量数据多尺度快速可视化技术。分析并提出提取土地利用现状数据多尺度信息时必须把数据的语义特征、空间结构特征以及图斑形状特征综合进行考虑的思想,构建了一种动静结合的矢量数据多尺度处理模型,实现了嵌入式设备上百兆级基于特征的矢量数据多尺度快速可视化技术。实验结果表明,该方法使得嵌入式GIS软件能够支持100MB以上矢量图的流畅显示,且整体平均渲染时间均不超过2秒,最恶劣情况下渲染时间不超过4秒,能满足大范围土地快速调查的需求。 (2)基于关键检测点识别的拓扑一致性面状图斑快速简化算法。列举分析了现有线要素化简算法存在的不足,通过研究顾及相邻地理实体空间关系的关键检测点识别方法,提出并实现了一种基于关键检测点识别的拓扑一致性快速化简算法。实验验证,该算法不仅能够避免简化后数据集中出现拓扑错误,实现一致性化简,而且在拓扑冲突检查时避免了不必要的拓扑一致性校验,提高了化简效率。 (3)基于改进G-H算法的超大图斑快速显示算法。针对在嵌入式设备上对超大尺寸图斑进行绘制时渲染耗时很长且由此易导致程序卡死这一问题,经实验揭示了该问题产生的原因是Windows CE、Windows Mobile中GDI的多边形图形绘制函数缺乏图形裁剪功能,提出引入多边形裁剪算法来解决该问题。详细分析了现有多边形裁剪算法在交点退化情况下算法失效原因,对G-H算法进行了改进,提出了一种顾及交点退化现象的多边形裁剪算法。实验验证,与以往算法相比,该算法不但在交点退化情况下能输出正确的裁剪结果,而且裁剪效率高。提出了基于多边形裁剪的超大尺寸图斑按需渲染策略,实验结果表明,与常规方法相比,所提渲染方法将嵌入式设备上超大尺寸图斑渲染效率提高了85%以上;当图斑放大到一定比例尺后渲染效率可提高94.5%以上,渲染时间甚至可从原先的几十秒降到1秒以内;将整幅矢量图显示效率提升了88%以上。 (4)自适应海量影像LOD快速可视化技术。针对资源有限的嵌入式设备难以处理、显示GB级影像数据这一问题,提出了一种海量影像金字塔构建优化算法,设计了LOD数据存储组织策略,提出了基于视相关LOD的遥感影像快速调度显示算法,实现了嵌入式设备上影像LOD快速显示。为了克服嵌入式设备存储卡自身硬件条件对嵌入式GIS软件系统实时处理影像数据能力的限制,提出了一种改进的自适应海量影像LOD表达处理算法。实验结果表明,改进方法使得嵌入式GIS软件能支持50GB以上影像数据快速渲染,渲染时间不超过1秒,且影像显示速度与原始影像大小无关。 (5)数字化动态变更外业调查与核查软件系统研发与应用。在上述关键技术研究成果的基础上研发了一套拥有自主知识产权的基于智能机的动态变更外业调查与核查系统,为快速获取准确、高现势性的土地基础数据提供了技术保障。从系统体系结构、框架结构、功能模块划分、总体工作流程这四个方面详细阐述了该系统设计方案。利用设计模式思想,基于Strategy、Command、Abstract Factory等模式进行软件开发,提高了嵌入式GIS框架的可复用性、可扩展性以及可维护性。 该系统已成功应用于全国土地调查、国家土地督察、全国年度变更调查、全国土地卫片执法检查,为我国制定相关调查方案和标准提供了重要参考。实际应用表明,实地调查效率较以往提高2-3倍,配合遥感数据,可以实现土地利用信息“月清季累”。系统已成为构建“数字城市”的动态空间地理基础设施,它的推广应用将会产生良好的经济效益和社会效益。
英文题目 RESEARCH ON TECHNOLOGY FOR MULTI-SCALE REPRESENTATION AND FAST RENDERING OF SPATIAL DATA IN EMBEDDED GIS
英文主题词 Embedded GIS, Land-use data, Remote sensing image, Multi-scale representation,LOD, Map generalization
英文摘要 The application of geographic information based on smart mobile terminals has become a development trend and has been a research focus in the field of geographic information science. With the increase of acquisition methods for geographic information, massive spatial data are continuously generated from various sources, such as smart mobile devices, UAVs, and remote sensing satellites. Currently, with the rapid development of computer hardware manufacturing industry, the performance of mobile terminals (embedded devices) has get a great improvement in computing speed, memory size and storage capacity. However, compared with spatial data volume, their computing resources and storage resources are still relatively limited, and still can not meet the demand for rapid processing of huge volume spatial data. The contradiction that the improvement rate of the performance of embedded devices can not match the growth rate of spatial data volume, also limits geographic information service quality. How to utilize the limited computing resources and storage resources of embedded devices to store, process and display massive spatial data is a very significant and urgent problem. Multi-scale representation for spatial data can be used to extract different levels of detail information, which is an effective method to realize rapid processing and rendering for spatial data on embedded devices. And in the overall strategy of ‘Digital China’ geo-spatial framework construction, the National Natural Science Foundation of China has put forward that multi-scale representation for spatial data is one of the core issues for geo-spatial databases technologies. Although the existing research has obtained significant research achievement, there are still some problems to be solved: most multi-scale information extraction technologies for vector data only consider spatial structure features, and do not take semantic features into account; image data storage and organization methods based on commercial database or distributed file system require high hardware configuration and complex management, which is not suitable for efficient storage and management of image data on embedded devices. In this dissertation, in order to achieve fast rendering for huge volume vector data and massive remote sensing image data on embedded devices, the key technologies for multi-scale representation and fast rendering of spatial data are studied. And a digital field investigation and verification system is developed to solve the integration problem of positioning information, digital map, and attribute information in land field survey. The main tasks of the dissertation are summarized as follows: (1) A fast multi-scale visualization technology for feature-based vector data. The thoughts that semantic features, spatial structure features and land parcels shape features must be considered when extracting multi-scale information of land-use data, is put forward. And a multi-scale processing model for vector data is constructed by using the offline and online generalization method. Then a solution for fast multi-scale visualization of feature-based vector data on embedded devices is proposed and implemented. The experimental results show that the proposed solution can enable the embedded GIS software to support fast scheduling and viewing for at least 100MB vector data with the average rendering time of less than 2 seconds, and even in the worst case, the rendering time is less than 4 senconds. (2) An algorithm for fast topological consistent simplification of face features based on key detection-point identification. The lack of traditional simplification algorithms for vector line is enumerated and analyzed. By studying the key detection-point identification strategy that takes spatial relationships into account, an improved DP algorithm based on key detection-point identification is proposed and implemented. The experimental results prove that the improved algorithm not only can guarantee planar topological consistent, but also can slightly reduce computational cost. (3) A fast rendering algorithm for super-size features based on the improved G-H algorithm. Aiming at the problem that the super-size features rendering on embedded devices is quite time consuming and can cause program exception phenomenon, the experiments reveal that the GDI polygon drawing function on Windows CE and Mobile platform is lack of graphics clipping function, thus it can not select the part of features which intersect with the display area to render. After analyzing the failure problem of the existing polygon clipping algorithms in intersection degenerate cases, an improved G-H polygon clipping algorithm is proposed and implemented. Compared with the previous algorithms, the proposed algorithm can not only resolve the degenerate cases, but also has a higher time efficiency. Then, an improved rendering strategy for vector data based on graphic clipping is put forward to accelerate the viewing speed. The experimental results show that compared with the original method, the improved method can boost rendering efficiency to more than 85% for super-size features, and even more than 94.5% when the display scale exceeds a certain value. The rendering time can even be reduced from tens of seconds to less than a second. And the entire map rendering efficiency can be increased at least 88%, achieving fluently display for vector map in the roaming, zooming process. (4) A LOD adaptive representation and fast visualization technology for massive image data. In order to resolve the conflict between the limited resources of embedded devices and the growing amount of massive image data to be shown, a solution for fast image data rendering on embedded devices is proposed and implemented. First, an improved algorithm for massive image pyramid construction is put forward, and a storage and organization strategy for tile data is designed. Then, a method, adopting technologies such as view-dependent levels of detail, target-tiles quick search and tiles seamless connection, is presented for fast scheduling and viewing of images. After that, to overcome the problem that embedded GIS software can not process images with more than 3GB, which is caused by the limitation of the memory card of embedded device, an improved adaptive solution for LOD processing and rendering of images is put forward. The experimental results show that the improved solution can enable the embedded GIS software to support fast scheduling and viewing for at least 50GB image data with the rendering time of less than a second, and the rendering speed does not depend on the image size. (5) Development and application of digital field investigation and verification system. Based on the above research results, a digital field investigation and verification system based on smart mobile devices with independent intellectual property rights is developed, which can provide technical support for quickly obtaining accurate geographic information data. The system design scheme is expounded from the aspects of system architecture, frame structure, function module division and overall work flow. Some design patterns, such as strategy, command and abstract factory, are adopted for improving the reusability, extensibility and maintainability of the embedded GIS framework. The system has been successfully applied to all kinds of land surveying business andication. can provide an important reference for the formulation of relevant investigation scheme and standards in our country. The practical application shows that the survey efficiency of the new method using the system is 2-3 times higher than that of the traditional method. It has become a digital and dynamic geographical information infrastructure, and will result in good economic and social benefits with its promotion and appl
学术讨论
主办单位时间地点报告人报告主题
东南大学 2011.06 中国南京 李 玲 New Method of Land Investigation Based on GPS/PDA
东南大学 2012.05 中国南京 李 玲 Image Compression Based on Wavelet
IEEE北京分会和中国人工智能学会 2012.10 中国杭州 李德毅 Artificial Intelligence
中国地质大学(武汉) 2015.11 中国武汉 刘耀林 地理空间优化技术与应用
东南大学 2016.05 中国南京 韦 钰 从AlphaGo看神经信息工程的发展
东南大学 2017.09 中国南京 芮 勇 人工智能系统的ABCD
东南大学 2013.09 中国南京 李 玲 图斑自动生成算法研究
东南大学 2014.11 中国南京 李 玲 基于空间数据的多尺度建模与网络渐进传输的外业数据采集系统的设计
东南大学 2017.10 中国南京 李 玲 一种顾及交点退化现象的多边形裁剪算法
     
学术会议
会议名称时间地点本人报告本人报告题目
全国地理信息科学博士生学术论坛 2014.11 中国南京 提高移动GIS复用性和高效性的软件设计与实现
测绘科学与技术全国博士生论坛 2015.11 中国武汉 城市公交优先关键技术研究与应用
     
代表作
论文名称
An algorithm for fast topological consistent simplication of face features
An algorithm supporting fast scheduling and viewing of RS images in embedded devices
 
答辩委员会组成信息
姓名职称导师类别工作单位是否主席备注
史照良 正高 教授 博导 江苏省国土资源厅副巡视员
闾国年 正高 教授 博导 南京师范大学
岳建平 正高 教授 博导 河海大学
程向红 正高 教授 博导 东南大学
潘树国 正高 教授 博导 东南大学
      
答辩秘书信息
姓名职称工作单位备注
王慧青 副高 副教授 东南大学