基于谱抠图的遥感图像滑坡半自动提取

摘要

In order to calculate the landslide areas from remote sensing image, a semi-automatic algorithm for extracting the landslide is proposed based on spectral matting. Matting Laplacian matrix is established to calculate feature vector, automatically determine the number of clusters, and hill-climbing algorithm is used for clustering image. Matting components are calculated based on feature vector and the user's data. The smoothing term is removed, and alpha is got. Experimental results show that the method can effectively extract the landslide, performs high accuracy and strong stability.%为统计遥感图像中滑坡区域的有效数据,提出基于谱抠图的遥感图像滑坡半自动提取方法.建立抠图拉普拉斯矩阵,计算特征向量,自动确定聚类数,利用爬山算法对图像聚类,根据特征向量和用户交互数据得到抠图成分,去除平滑项,得到前景透明度.实验结果证明该方法能够有效提取滑坡信息,准确率高,稳定性强.