基于高分辨率遥感影像建筑物提取研究

发布时间：2018-07-11 21:46

本文选题：面向对象 + 灰度阈值分割　；参考：《中南大学》2013年硕士论文

【摘要】：在地理空间数据库中建筑物是核心地形要素之一,同时也是城市环境中不可缺少的重要组成部分以及人类活动的重要聚居地。而随着社会的发展,建筑用地不断的发生变化,加快空间数据库的更新也变得非常重要。遥感技术的发展,使得人类精确识别和提取地物信息,自动更新GIS数据库成为可能。因此,本文着重围绕如何从高分辨率遥感影像上提取建筑物轮廓展开研究,研究的主要内容如下： (1)利用影像中波段数值运算的原理与绿波段中植被的反射率比较大的特性,采用绿波段与红波段的差值运算方法,消减其他地物的亮度值,增加植被亮度值,经过形态学修复之后,采用简单灰度阈值分割的算法提取出植被,再进行掩膜处理,消除植被对后续建筑物提取的干扰；最后,采用相同分辨率的不同区域的遥感影像数据验证该植被提取方法的通用性。 (2)详细分析多尺度分割算法原理和分割尺度的确定方法。采用多尺度分割算法和K邻近的分类方法将去除植被的影像进行分割和分类识别,得出建筑物轮廓；最后设计了三个个对比试验：1)仅仅利用灰度特征提取建筑物轮廓的结果与本研究采用方法的结果相比较,分析提取精度。2)本研究提取方法的结果与传统的基于像元的分类提取结果进行对比,结果表明：利用面向对象的思想对本研究中的高分辨率遥感影像进行分类的方法具有明显的优越性,提取精度明显提高。3)在相同分割条件下,对比分析采用知识规则分类提取结果与采用本研究方法提取结果,分析提取精度。 (3)文章提出了一种基于数字化的方法对于建筑物提取的结果中漏分、错分的现象进行后处理,即使用3种不同的数字化方法(手扶跟踪数字化,自动跟踪数字化,GIS数据叠加匹配)分别对轮廓破坏建筑物轮廓进行了修复处理。为了加快提取速度,本文提出将提取的建筑物轮廓与GIS数据库中有的相同地区矢量数据匹配叠加,快速自动修复破坏了的建筑物轮廓。三种方法都是通过c#和arcengien平台开发实现。
[Abstract]:In the geospatial database, the building is one of the core terrain elements, and it is also an indispensable part of the urban environment and an important settlement of human activities. With the development of society, the land for building is changing constantly, so it is very important to speed up the updating of spatial database. With the development of remote sensing technology, it is possible to identify and extract the feature information accurately and update GIS database automatically. Therefore, this paper focuses on how to extract building contours from high-resolution remote sensing images. The main contents of the research are as follows: (1) by using the principle of wave band numerical operation in image and the characteristics of high reflectivity of vegetation in green band, the difference operation method between green band and red band is adopted to reduce the brightness value of other ground objects. Increase the brightness value of vegetation, after morphological repair, the simple gray threshold segmentation algorithm is used to extract vegetation, and then mask processing to eliminate the vegetation interference to the subsequent building extraction; finally, Remote sensing image data of different regions with the same resolution are used to verify the generality of the vegetation extraction method. (2) the principle of multi-scale segmentation algorithm and the method of determining segmentation scale are analyzed in detail. Multi-scale segmentation algorithm and K-neighborhood classification method are used to segment and identify the vegetation removal image, and the building contour is obtained. Finally, we design three contrast experiments: 1) the results of extracting the building contours using only gray features are compared with the results of the method used in this study. Analysis and extraction accuracy .2) the results of this method are compared with those of traditional pixel based classification. The results show that the method of classifying the high resolution remote sensing images in this study has obvious advantages, and the precision of extraction is improved by 3. 3) under the same segmentation conditions, the method can be used to classify the high resolution remote sensing images. Comparing the results of classification with knowledge rules and the results of this study, the paper analyzes the accuracy of extraction. (3) this paper proposes a digitized method for the missing points in the results of building extraction. Three different digitization methods (walking tracking digitization, automatic tracking digital GIS data superposition matching) were used to repair the damaged building contour. In order to speed up the extraction speed, this paper proposes to match and superposition the extracted building contour with the same area vector data in GIS database, so that the damaged building contour can be repaired quickly and automatically. All three methods are developed on the platform of c# and arcengien.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P208

【参考文献】