【摘要】 由于高光谱图像的丰度特性,盲源分离算法不能直接用于高光谱图像解混。同时,在解混过程中用梯度算法对目标函数进行优化时易陷入局部最优。为此,本文提出了一种基于多目标蝙蝠优化算法的高光谱图像解混算法。该算法将高光谱图像模型中存在的丰度非负约束及丰度和为一约束作为解混的两个目标函数,将解混问题转化为对目标函数的优化问题,同时引入多目标蝙蝠优化算法来求解,从而实现高光谱图像解混。实验结果表明,本文算法能有效解决上述问题,并且当改变图像的信噪比、像元纯度和像素数时,观察光谱角距离与均方根误差值的变化情况,与其它解混算法相比,本文算法具有更高的解混精度和很好的抗噪性,在像元纯度很低的情况下也有很好的性能。更多还原

【Abstract】 For the abundance speciality,blind source separation algorithm is difficult to be used for the hyperspectral unmixing directly.At the same time,traditional gradient algorithm is easy to fall into the local extremum when it optimize the objective function in unmixing process.For this purpose,we propose a hyperspectral unmixing algorithm based on multi-objective bat optimization algorithm in this paper.This algorithm makes the abundance non-negative constraint and abundance sum-to-one constraint as objective functions for unmixing.Thus,unmixing problem is transformed into a multi-objective function optimization problem.At the same time,we introduce the multi-objective bat optimization algorithm to optimize the objective function,thus realize the hyperspectral unmixing.The experimental results show that the hyperspectral unmixing algorithm based on multi objective bat optimization can solve the above problems effectively,in addition,when the signal-to-noise ratio,purity of pixel and the number of pixels in image are changed,the changes of spectral angel distance and root mean square error are observed.We can see that compared with other unmixing algorithms,the new algorithm has higher decomposition accuracy,and its anti-noise interference ability is strong.It can also be used at mixed pixel decomposition with very low purity.更多还原