提取建筑物屋顶是建筑物三维重建的一个重要步骤，为了更精确的提取建筑物屋顶，本文融合了高分辨率的光学和SAR图像。利用SAR图像中建筑物的叠掩来确定光学图像中的建筑物，从而在光学图像上确定一个包含建筑物屋顶小窗口，利用ISODATA分类方法对小窗口进行了分类来获取建筑物屋顶类别或屋顶轮廓，最后计算得到屋顶的规则轮廓。从计算结果上看该方法是可行的。

The extraction of building’s roof is an important step in 3D reconstruction of building. In order to more precisely extract the roof information, both high-resolution optical and SAR images were fused in this paper. Layerover of building in SAR image was used to determine that building on optical image, and thereafter a small window was setup including building’s roof. The ISODATA classification method was used to obtain an outline of building’s roof for the small window. Finally, the regular outline of the roof was computed. The experiment proved that this method was feasible.