中文摘要：

针对高空间分辨率卫星图像上建筑物的特点，提出了一套半自动、快速简单的建筑物二维轮廓信息提取技术流程。首先，利用最大同质性近邻滤波算法（MHN）对图像进行预处理；然后，利用基于数学形态学经典算法（腐蚀、膨胀、开操作及闭操作等）重建的微分形态学断面方法（DMP）进行建筑物二维轮廓信息提取，并在结构算子的步长选择方面对方法进行了初步改进；最后，结合区域标识和面积阈值分割等方法进行了后处理操作。以北京师范大学校园QuickBird图像为例的试验结果显示，该方法不仅对规则建筑物的提取效果良好，还可以有效地提取出形状复杂的建筑物轮廓信息。

英文摘要：

In view of characteristics of high - resolution satellite images, this paper has designed a set of technological schemes for semiautomatic rapid and easy extraction of building 2 - D rooftop information. Firstly image preprocessing was performed by using the Maximum Homogeneity Neighbor Filter Method. Then Differential Morphological Profile Method was employed which is based on the reconstruction of classical morphological algorithms such as erosion, dilation as well as open and close operators. Some preliminary improvements were made in choosing the step of structure operator. The image after - processing was made based on the combination of Label Region and Threshold Segmentation. Finally the QuickBird image was chosen which covered the campus of Beijing Normal University as the study area because of its dense man - made buildings. The result shows that this set of technological schemes not only has a good extraction effect for regular buildings, but also can extract the outline of the complex building effectively.