激光与光电子学进展, 2020, 57 (6): 061020, 网络出版: 2020-03-06
一种改进的高光谱解混非负矩阵分解初始化方法 下载: 976次
Improved Hyperspectral Unmixed Initialization Method Based on Non-Negative Matrix Factorization
图像处理 高光谱影像 非负矩阵分解 空间特征 光谱特征 imaging processing hyperspectral image non-negative matrix factorization spatial characteristics spectral characteristics
摘要
提出了一种结合欧氏距离和光谱信息散度的改进的高光谱解混非负矩阵分解(NMF)初始化方法(IISSF)。在初始化基础上,结合标准NMF算法和分块NMF算法进行平行对比实验。结果表明,在合成影像实验中,在信噪比为20 dB~50 dB范围内,经过IISSF初始化后的分块NMF算法获取的结果要优于其他方法;且其在真实影像实验中获取的端元光谱与真实影像端元光谱之间具有最小的平均光谱角差值,即0.1812 ;其重构影像与真实影像之间的均方根误差值最小,为0.007。
Abstract
An improved hyperspectral unmixed initialization method (IISSF) based on non-negative matrix factorization (NMF) combining Euclidean distance and spectral information divergence is proposed. On the basis of initialization, a parallel comparison experiment is performed in combination with the standard NMF algorithm and the block NMF algorithm. The results show that, in the synthetic image experiment, the block NMF algorithm after IISSF initialization is better than other methods in the signal-to-noise ratio range from 20 dB to 50 dB. There is a minimum average spectral angular difference between the end-member spectrum obtained in the real image experiment and the reality image endmember spectra, i.e., 0.1812. The root mean square error between the reconstructed image and the real image is the smallest, i.e., 0.007.
黄鹏飞, 孔祥兵, 景海涛. 一种改进的高光谱解混非负矩阵分解初始化方法[J]. 激光与光电子学进展, 2020, 57(6): 061020. Pengfei Huang, Xiangbing Kong, Haitao Jing. Improved Hyperspectral Unmixed Initialization Method Based on Non-Negative Matrix Factorization[J]. Laser & Optoelectronics Progress, 2020, 57(6): 061020.