面向高光谱图像的高斯-稀疏子空间聚类算法

Gaussian-sparse subspace clustering algorithm for hyper spectral imagery

稀疏子空间聚类算法聚类图像是基于谱聚类实现的,谱聚类的关键是构造图的相似度矩阵,在稀疏子空间算法的基础上,提出一种利用稀疏系数矩阵与经过PCA处理后像素的高斯相似度来构造图的相似度矩阵的聚类算法。实验结果表明,提出的聚类算法充分提高了高光谱图像的聚类精度。

The sparse subspace clustering algorithm is based on spectral clustering.The key of spectral clustering is to construct the similarity matrix of the pixels.Based on the sparse subspace algorithm,a method is proposed which uses sparse coefficient matrix and Gaussian-similarity of hyper spectral pixels that are processed through PCA.Experimental results show that the proposed clustering algorithm fully improves the clustering accuracy of hyper spectral images.