高分辨率遥感影像的建筑物自动提取

Buildings automatic extraction from high resolution RS images

文章结合结构图分析方法,提出一种针对高分辨率遥感影像中建筑物自动提取方法。该方法利用几何限制对线基元进行初连接,解决由于边缘检测无法完整提取建筑外边缘问题;通过构造结构图确立线基元之间连接关系,搜索结构图中闭合路径,建立高级别特征,对过于散列的线基元进行整合;在此基础上,综合考虑建筑物的几何结构、光谱特征等图像信息筛选闭合路径,实现对建筑物的提取。最后通过对天绘一号高分辨率遥感影像进行建筑物自动提取验证了该方法的有效性。

Combing with the analysis method of Structure Graph, an approach of extracting buildings from high resolution remote sensing images was proposed in this paper. Firstly, the line primitives were joined by geometric conditions, in order to solve the problem of incomplete detecting of building edges. Secondly, high-level features were established to integrate the discrete line primitives, by construc- ting structure graph to identify the relationship among primitive lines, and search closed paths. Then, ac- cording to multiple types of image features such as geometric structures and spectral characteristics, closed paths were filtered to extract buildings. Finally, experiments applying remote sensing images of TH-01 were carried out and the result verified the validity of the approach.