【作者】 段磊；

【导师】 刘勇；

【作者基本信息】 兰州大学 ， 地理学·地图学与地理信息系统， 2016， 硕士

【摘要】 乡村居民地是一类重要的基础地理信息,快速有效地监测乡村居民地的时空格局,不仅对地理国情普查具有重要作用,而且对于其区域可持续发展和乡村城市化具有积极意义。随着遥感技术的飞速发展,其覆盖范围广、获取速度快、空间分辨率高的特点有利于快速准确地掌握乡村居民地空间信息。当前研究高分辨率遥感影像居民地这一类复合地物信息提取的方法,大多采用基于像元方法分析居民地的结构纹理(边缘检测和角点分析),其方法存在诸多的局限性。探索新的高分辨率影像复合地理对象的信息提取方法是必然趋势。在遥感大数据的背景下,采用本体建模和基于对象影像分析方法相结合的方式,探索自动/半自动提取乡村居民地这一类复合地物信息,对遥感自动化信息提取具有积极意义。本体建模其实质就是建立对象所涉及知识的逻辑模型,其中知识主要指对象的概念、概念间的关系、属性和约束条件[1,2]。本文以宁夏回族自治区中卫市沙坡头区河滩村及其周边为研究区,采用WorldViewⅡ为数据源,结合本体建模与基于对象影像分析方法提取高分辨率影像乡村居民地信息,并着力于提高乡村居民地信息的完整性和准确性。首先本文研究了乡村居民地本体概念模型的建立方法,其次探讨了如何利用基于对象影像分析方法形式化本体模型,最后采用不同的信息提取方法进行了对比分析,并验证了基于本体模型提取乡村居民地信息的有效性。通过以上研究,本文得到了如下结论:(1)提出高分辨率遥感影像乡村居民地认知框架,建立乡村居民地本体的概念化模型,使乡村居民地在高分影像中的描述具有一定的明确性和共享性。并证明了根据本体模型提取乡村居民地信息的研究思路是可行的。(2)高分辨率遥感影像中乡村居民地是一类复合地物,其光谱信息十分复杂,利用基于对象影像分析技术与本体建模相结合的方法,对高分辨率复合地物信息提取提供了一种有效的途径,具有积极意义。(3)构建乡村居民地本体模型时需要注意利用共有的知识构建本体模型,即遵守一致性原则。一致性原则指构建本体的概念和对象的内涵要一致,本体模型中公理和非形式化的相关概念要一致,否则建立的本体模型其共享性低。(4)OBIA除了可以利用光谱和纹理信息外,还可以添加几何和上下文信息用于信息提取,利用多种类型的分类特征更加精确地描述乡村居民地本体模型,有效地提高了乡村居民地信息提取的准确性和完整性。(5)OBIA方法中影像分割质量对信息提取的精度具有较大影响。影像分割对象与信息提取对象越匹配,信息提取的结果越准确。基于PSE-NSR-ED2的最优分割参数选择方法将几何和误差代数误差同时考虑,有效地提高了影像分割质量。(6)利用OBIA方法形式化影像乡村居民地本体模型,主要包括4个方面的内容:乡村居民地本体模型概念集的OBIA参数表示;本体模型关系集的OBIA参数表示;本体模型属性集的OBIA参数表示和本体模型公理集的OBIA参数表示。参数表示的过程在基于对象影像分析中分别对应于建立分类体系、确定关系集与属性集的分类特征、建立规则集。更多还原

【Abstract】 Rural residential area(RRA) was significant basic geographic information. Monitoring the time-spatial coordination of RRA fleetly and efficiently, not only played a momentous role in census of national geographic conditions, but also was of great positive significance to regional sustainable development. With the rapid development of remote sensing, it had many virtues, such as wide coverage, easily accessible and high spatial resolution. Therefore we could grasp the spatial information of RRA rapidly and accurately.At present, the research on information extraction of composite feature in high-resolution remote sensing images, mostly adopted the method to analysis structure-texture of RRA based on pixels, but the method had a lot of limitations. Therefore, exploring new ways to extract information of composite feature was inevitable trend. Under the background of big data on remote sensing, geographic information ontology modeling was an important content of study on automatic/semiautomatic land cover types identification from remote sensing image, and played an important role in information extraction of composite feature in high resolution remote sensing image. Gruninger etc.(1995) considered that the essence of ontology modeling was establishing logical model based on objects’ knowledge, such as concepts, relations, attributes, constraint conditions, etc.In this paper, using the WorldView Ⅱ satellite image of the Hetan Villiage and surrounding, in Shapotou District, Zhongwei County, Ningxia Province, China, It discussed the method about RRA information extraction based on ontology modeling and object-based image analysis, and focused on improving the integrity and accuracy. At first, it studied the way to build ontology model of RRA. Secondly, it discussed how to formal ontology model using OBIA. Finally, it used different ways to extract RRA information and verified that ontology modeling method was effective.The result indicated that,(1) putted forward the cognitive framework of RRA in high resolution imagery and built the RRA conceptual model to extract the information of rural residential area was a feasible idea.(2) RRA was a kind of complex feature. Its spectral information was very complex. Therefore, taking advantage of OBIA and ontology modeling was an effective way to extract complex feature information from high resolution remote sensing images. This method had positive significance.(3) At the time of building RRA ontology model, we should pay attention to the share knowledge used to construct ontology model.(4) Using an object-based image analysis method to formalize the conceptual model of rural residential, it could take advantage of the spectral characteristics, geometric feature, texture feature, spatial relationships and context information. So the RRA ontology model built before was more accurate. It improved accuracy and completeness of the information extraction.(5) Image segmentation quality serious affected the precision of result. More accurate segmentation led to more precision result. “PSE-NSR-ED2” was a better way to select the optimal segmentation result. It considered the both algebra difference and geometry difference.(6) Formalizing ontology model of RRA using OBIA included several steps: expressing the sets of concepts, relations, attributes and axioms with OBIA. They respectively represented “establish classification system”, “determine classification characteristics”, and “build rule sets”.更多还原