基于改进分水岭变换的遥感图像建筑物提取

遥感图像的分割是遥感信息提取与目标识别的基础和关键。以高分辨率城镇地区遥感图像为研究对象,提出一种基于全局阈值的多级分水岭算法,用于遥感图像的分割。该算法通过引入差异度函数,在执行传统分水岭算法的过程中对图像中存在的噪声区域进行修正,并通过使用全局阈值有效的控制欠分割问题。首先基于全局阈值的多级分水岭算法对高分辨率遥感图像进行初始分割,然后综合利用分割对象的颜色和形状特征信息,进行区域合并和梯度边缘提取,得到最终的建筑物提取结果。实验结果表明,所提出的基于全局阈值的多级分水岭算法较好地避免了过分割和欠分割现象,结合区域合并和梯度提取,能够快速准确地对城镇遥感图像中的建筑物进行提取。

BUILDING EXTRACTION FROM REMOTE SENSING IMAGE USING IMPROVED WATERSHED TRANSFORM

The segmentation of remote sensing image is the basis and key of information extraction and target identification.The paper uses remote sensing image of high resolution in urban areas as the research object and does study on the segmentation of buildings in images.We proposed a multistage watershed algorithm based on global threshold.The algorithm decreases the noise of images in the implement process of watershed algorithm through the introduction of a differential function and overcomes the problem of ove...