Hyperspectral image unmixing method and system based on unsupervised learning

Abstract

The invention discloses a hyperspectral image unmixing method and system based on unsupervised learning. The method comprises the following steps: acquiring a hyperspectral image; performing super-pixel segmentation on the hyperspectral image by adopting a clustering algorithm to obtain a plurality of super-pixel regions; carrying out homogeneity analysis on the super-pixel region to obtain a homogeneous super-pixel region and a non-homogeneous super-pixel region; respectively processing the homogeneous superpixel region and the non-homogeneous superpixel region to obtain the characteristic representation of the hyperspectral image; and performing non-negative matrix factorization on the feature representation of the hyperspectral image under a constraint condition to obtain the spectrum feature and abundance of each end member in the hyperspectral image. According to the method, the complexity of a demixing algorithm can be reduced, the non-homogeneous spectral feature information contained in the hyperspectral image is fully mined, and the demixing precision is remarkably improved.

Images