论文研究-基于互信息加权的最小二乘法丰度反演 .pdf

反演，选择互信息矩阵作为加权矩阵，从熵的角度反映了不同波段间的相关性。同时，在丰10

果得到提升，验证了该算法的可行性。

关键词：信息与信号处理；高光谱解混；丰度反演；最小二乘算法；互信息；波段选择

Abstract: Abundance inversion based on Least squares algorithm is more commonly used in the

hyperspectral mixed pixels unmixing. Traditional least squares error problem in abundance

inversion process, equally weighted all bands, without considering the contributions of each band

are different, which may not be accurate for the hyperspectral data. In order to highlight the

distinctness between the bands and obtain more accurate abundance of mixed pixels, this paper 25

uses the least squares error algorithm which is based on weighted matrix for the abundance

inversion. Abundance inversion based on mutual information-weighted least squares error

algorithm is presented in this paper, mutual information from the perspective of entropy to reflect

the correlation between different bands. Also it uses band selection technology in abundance

inversion to reduce the complexity of data processing. Compared with the existing weighted 30

度约束条件不同，最小二乘法可分为：无约束最小二乘法、“和为 1”最小二乘法、“非负”40

法进行的。