分组寻优的多端元高光谱图像解混方法

Multi-endmember hysperspectral image unmixing based on group optimization

为了对高光谱图像中出现的混合像元进行分解,在分层解混方法的基础上从一个新的角度出发,提出了一种新的高光谱图像解混方法。该方法根据整体和部分之间的关系,首先利用各种地物的端元组对混合像元分别进行解混,然后利用根均方误差(RMSE)选出每个端元组中反演误差最小的那个端元。该方法可以降低计算的复杂度,有效地抑制噪声的影响,迭代次数减少了1900次。

In order to decompose the mixed pixel in a hyperspectral image,in this paper,a new hyperspectral image unmixing method is proposed based on hierarchical multi-end hyperspectral image unmixing.According to the relationship of the whole and the part,the method first uses the endmembers of each land-cover class,and then uses the root mean square error(RMSE)to select the endmember having the minimum inversion error in each land-cover class.This method can reduce the difficulity of computation,effectively restrain the influence of noise,and the number of iterations has been reduced by 1900.