Hyperspectral Remote Sensing Inversion of Mineral Abundance Based on Sparse Unmixing Method

The task of mineral abundance inversion using hyperspectral images is a promising challenge. This study presents several existing state-of-the-art sparse unmixing algorithms that combine spectral and spatial information from hyperspectral images. They show good performance in both simulation and real hyperspectral datasets. In particular, the spectral information and spatial structure information in hyperspectral images can be more fully utilized by introducing the superpixel segmentation algorithm. Taking the Cuprite dataset as an example, which is a real mining dataset, experiments indicate that the sparse unmixing algorithm achieves satisfactory results on this dataset.