一种基于光谱距离约束的非负矩阵分解算法

A non-negative matrix factorization algorithm based on spectral distance constrained

在处理混合像元分解问题中,非负矩阵分解算法是一种比较热门的算法,因其算法模型与线性光谱模型有着相似性和解出的结果符合现实要求而被广泛应用。现有的非负矩阵分解方法在针对端元之间差异性的同时,并没有考虑到有部分端元可能会有相似的光谱特征的存在。因此,提出一种基于光谱距离约束的非负矩阵分解算法,该方法在对图像中所有类别进行解混的同时,既考虑到不同物质端元之间的可分性,也考虑到具有相似光谱特征物质之间的端元相似性。实验结果表明了所提方法的有效性。

The non-negative matrix factorization （NMF） algorithm, one of the popular decomposition algorithms of mixed pixels, has been widely used because it is similar to linear spectral model and its results are accord with the reality requirements. The existing NMF algorithm deals with unmixing problems according to the differences among all the endmembers in the imagery, at the same time, it does not take into account that some endmembers may have similar spectral characteristics. Therefore, a kind of spectral distance constrained non-negative matrix factorization is proposed. This method not only takes into account the separability of different endmembers, but also similarity of endmembers which have similar spectral features. Experiments show the effectiveness of the proposed method.