改进的MVC-NMF算法在高光谱图像解混中的应用

高光谱图像中混合像元的存在不仅影响了基于遥感影像的地物识别和分类精度,而且已经成为遥感科学向定量化发展的主要障碍。MVC-NMF混合像元解混算法不需要假定纯像元的存在,并且在自动提取端元的同时,能够获取每种端元所对应的丰度图,然而该方法收敛速度较慢、效率较低,文中提出利用BBA(Barzilai-Borwein Algorithm)算法来实现MVC-NMF算法的迭代更新,从而提高了算法的效率。实验结果表明:该方法在相同迭代次数条件下比MVC-NMF算法解混的结果好、效率高。

The Application of Advanced MVC-NMF in the Hyperspectral Image Unmixing

The mixels in the hypersepectral images not only influence the accuracy of target detection and classification,but also greatly hinder the development of quantitative remote sensing. The MVC-NMF algorithm is not based on the hypothesis of the pure pixels'' existence,and can get the abundance maps as soon as the endmembers are all extracted,but it converges very slowly,and has low efficiency. We propose to use the NMF- GPSR-BB algorithm to update the MVC-NMF algorithm. Experimental results demonstrate that the proposed scheme for decomposition of mixels outperforms the MVC-NMF algorithm under the same number iteration condition.