基于最小体积约束的非负矩阵分解模型的高光谱解混算法探究

高光谱遥感图像中,遥感影像的分类精度和地物识别会因混合像元的存在而受到影响,从而限制了遥感科学向定量化发展.基于最小体积约束的非负矩阵分解方法,不仅不需要假定纯像元的存在,而且在自动提取端元的同时获取对应的丰度图,这种非监督的光谱解混技术克服了传统方法的限制条件,为高光谱图像中混合像元问题的解决提供了新的思路和方法.

Research on Algorithms for Non-negative Matrix Factorization Method Based on Minimum Volume Constraint in Hyperspectral Image Unmixing

The existence of the mixed pixels in the hyperspectral remote sensing images affects the classification accuracy and the object recognition of the remote sensing images,which has become the main obstacle of developing quantitatively the remote sensing science. Non-negative matrix factorization method based on minimum volume constraint（ MVC-NMF） not only does not need to assume the existence of pure pixels,but also can obtain the corresponding abundance map while automatically extracting the endmember. This unsupervised spectral unmixing technique overcomes the limitations of traditional methods,which provides new ideas and methods for solving the mixed pixel problems in the hyperspectral images. The theory and application of this algorithm have become a hotspot of research in recent years.