基于区域分割合并的建筑物半自动提取方法  

Semi-automatic Extraction of Building Based on Region Segmenting and Merging

作　　者：周俊 晏非 孙曼 

机构地区：[1]西安测绘信息技术总站,陕西西安710054

出　　处：《海洋测绘》2005年第1期58-60,共3页Hydrographic Surveying and Charting

摘　　要：提出了一种遥感图像半自动提取建筑物的方法。首先通过对建筑物边缘进行折线逼近和感知编组,可以得到其顶部大体结构。然后按照"二叉树"的数据结构,利用"超线"对建筑物影像进行区域分割。最后通过对分割后的区域进行相似性的统计并按照"二叉树"记录的分割顺序进行生长合并,从而得到了屋顶的边缘,实现了对平顶直角建筑物的提取。该方法能够充分利用区域的性质进行合并,稳定精确地提取建筑物。The paper presents a method for semi-automatic extractin of building . We arrange the line segments of (building's) edges to reconstruct the structure of the building roof by means of line closing and perceptual organization method first. And then, a binary space partitioning tree is adopted as an efficient data structure to aid recursive partitioning of a region with hyperline. When the image of the building is segmented into statistically homogeneous sub-regions, we can merge the regions by using pre-defined homogeneous predicates. Through the above steps, the roof of the building can be extracted accurately and steadily.

关 键 词：像片判读 建筑物提取 区域分割 超线 二叉树 

分 类 号：TB22[天文地球—大地测量学与测量工程]