高分辨率遥感影像中建筑物外形提取方法研究.pdf

摘 要
随着商用遥感卫星的成功发射，高分辨率遥感图像资源越来越丰富。由
于具有信息量大、获取迅速、时效性好、合理的经济成本等特点，它在国民
经济活动中扮演着越来越重要的角色，因此高分辨率遥感图像中的地物信息
提取技术成为研究的热点。建筑物是城市模型中最主要的基元，精确地构建
建筑物外形在许多领域中有重要的意义。
本文从典型的建筑物外形出发，对高分辨率遥感影像建筑物外形半自动
提取方法进行了研究。主要采用了两种方法进行提取：一种是根据建筑物外
形的边缘进行提取，通过对图像的去噪，区域生长等图像处理方法排除了大
部分的非建筑物外形边缘，然后通过边缘检测、线段规划方法最终提取出建
筑物外形；另一种是根据建筑物的特征进行提取，在多尺度分割技术的支持
下，将图像分割成许多多边形对象，依据建筑物的各种特征对各个多边形对
象进行筛选和提取，最后采用数学形态学处理技术将提取出来的对象外形进
行优化，得到更加真实的建筑物外形。试验证明，这两种方法对于提取建筑
物外形都具有很好的效果。

关键词：建筑物外形提取；图像分割；建筑物外形优化；Hough 变换；图像
分类
I
ABSTRACT
With the commercial remote sensing satellites were launched successfully,
high-resolution remote sensing imagery has become progressively more
abundant. Because of its large amount of information, quick access, timeliness,
and a reasonable economic cost, It plays an increasingly important role in the
national economy. So the technology of extracting information of the remote
sensing images once become a hot research again. Building is the most
important element of the cities, therefore, constructing the precise outline of
buildings is very important in many projects.
This paper studies the typical buildings in the high-resolution remote
sensing images. There are two ways to extract the outline of buildings: one is
based on the edge of the buildings, through the removal noise, the growth of the
region, and other image processing methods to exclude the most of the
non-pixels which belong to the building edge , And then through the edge
detection, linear programming method to extract the final shape of buildings.
The other is based on the characteristics of buildings, in the support of
multi-scale partition technical , the image will be divided into a number of
polygon objects, according to various characteristics of the buildings , every
polygon object will be screened and extracted, and then adopts mathematica