高光谱遥感图像非线性解混研究综述

Review of nonlinear unmixing for hyperspectral remote sensing imagery

介绍了近年来非线性光谱解混方法的发展状况,主要包括矿物沙地地区的紧密混合模型和植被覆盖区域的多层次混合模型,以及基于这些模型的非线性解混算法和利用核函数、流形学习等方法的数据驱动非线性光谱解混算法及非线性探测算法.最后分析总结了现有非线性解混模型与算法的优势与缺陷及未来的研究趋势.

The development of non-linear spectral unmixing methods in recent years is introduced. There are mainly two types of models. One is the close-mixing model of mineral sand area and the other is multi-level mixing model of vegetation coverage area. The data-driven nonlinear spectral unmixing al- gorithms such as kernel methods and manifold learning are presented. Both advantages and disadvanta- ges of these models and algorithms are summarized and the future research trends are analyzed.