高光谱图像非线性解混方法的研究进展

由于空间分辨率的限制,高光谱遥感图像中存在大量混合像元,对混合像元的解混是实现地物精确分类和识别的前提。与传统的线性解混方法相比,非线性解混方法在寻找组成混合像元的端元以及每个端元的丰度时具有较高的精度。分析了光谱非线性混合的原理,总结了近年来提出的非线性解混算法,重点对双线性模型、神经网络、基于核函数的非线性解混算法以及基于流形学习的非线性解混算法进行了介绍和分析。最后总结了混合像元非线性解混未来发展的趋势。

Advances in Nonlinear Spectral Unmixing of Hyperspectral Images

Due to the limitation of spatial resolution,there are lots of mixed pixels in spaceborne hyperspectral images.Spectral unmixing of hyperspectral images is an important premise for accurate terrain classification and identification.Compared with traditional spectral unmixing techniques based on the linear mixing model,nonlinear spectral unmixing techniques has better performance in finding endmembers and their abundances.The principle of nonlinear spectral mixture is analysed,and nonlinear unmixing algorithms increased in recent years are summarized.This paper emphatically introduces bilinear model,neural networks,nonlinear spectral decomposing based on kernel function and manifold learning.Some future directions of research are introduced.