基于分水岭变换的单幅高分辨率SAR图像建筑物检测方法

提出了一种从单幅高分辨率SAR图像中检测建筑物的算法框架。该框架主要以应用标记的分水岭变换为基础,针对SAR图像中建筑物所具有的强回波特性与典型的形状特征,主要采用CFAR检测和文中提出的方向相关分析方法得到标记图像,利用最小强制技术和标记图像修改原始图像的梯度图,对修改后的梯度图作分水岭变换得到建筑物目标的边界轮廓。该方法能够引入建筑物目标的特性同时克服分水岭变换固有的过分割缺陷。文中对不同场景的高分辨率SAR图像进行了实验,实验结果表明,即使在建筑物分布密集的情况下,本文算法也能正确完整地检测出绝大多数目标,检测率高而虚警率低。

Building Detection from a Single High-resolution SAR Image Using Watershed Transform

This paper deals with detecting buildings froma single high-resolution synthetic aperture radar (SAR) image.A building detection framework based on the marker-controlled watershed transform was proposed.In this framework,the strong radar backscatter energy and shape features of buildings in SAR images were introduced as markers,which were extracted mainly by a constant false alarmrate (CFAR) detector and a method of shape analysis called direction correlation analysis.Combined with the markers,the minima impositiontechnique was implemented on the gradient image of the original SAR image in order to overcome the oversegmentation drawback of a directly-computed watershed.The building boundaries were then obtained by computing the watersheds of the modified gradient image.We applied the proposed method to high-resolution SAR images of different scenes.The results reveal that this method has a high detection rate and alow false alarmrate even if the buildings are densely distributed.