Improved hyperspectral unmixing initialization method and its applicability analysis

doi:10.13474/j.cnki.11-2246.2022.0233

Abstract

Abstract: There are a large number of mixed pixels in hyperspectral images, which greatly limit the quantitative application of hyperspectral images, and it is especially important to perform pixel unmixing efficiently and accurately. The initialization of the endmember matrix, the cost function of the algorithm itself and the iterative rules of the algorithm often result in different end-element spectra and end-batch abundances. Under different conditions, it is especially important to select appropriate initialization methods, cost functions and iterative rules to make the hyperspectral unmixing results better. In this paper, a new block initialization method based on Euclidean distance and spectral information divergence is improved. The improved block initialization method is superior to other initialization methods in the case of low to medium SNR. At the same time, a lot of experiments are carried out on the two aspects of initialization and algorithm itself. The results show that:①The block initialization is better than the global initialization.②The gradient iterative NMF algorithm is faster than the multiplicative iterative NMF algorithm, but easy to fall into local minimum.③Block multiplicative iterative NMF algorithm can obtain better endmember abundance information than standard multiplicative iterative NMF algorithm;④Block gradient iterative NMF algorithm is not applicable to the spectral unmixing process after random initialization.

Key words: hyperspectral image, non-negative matrix factorization, spectral unmixing, initialization

CLC Number:

P237

WEI Lei, DING Laizhong, WANG Wenjie, GAO Yantao, HUANG Pengfei, LI Chunyi, ZHANG Yongjie. Improved hyperspectral unmixing initialization method and its applicability analysis[J]. Bulletin of Surveying and Mapping, 2022, 0(8): 61-67.

References

[1] PASCUAL-MONTANO A,CARAZO J M,KOCHI K,et al. Nonsmooth nonnegative matrix factorization (nsNMF)[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2006,28(3):403-415.
[2] HOYER P O. Non-negative matrix factorization with sparseness constraints[J]. Journal of Machine Learning Research,2004,5:1457-1469.
[3] WANG X Y,ZHONG Y F,ZHANG L P,et al. Spatial group sparsity regularized nonnegative matrix factorization for hyperspectral unmixing[J]. IEEE Transactions on Geoscience and Remote Sensing,2017,55(11):6287-6304.
[4] LEE D D,SEUNG H S. Algorithms for nonnegative matrix factorization[J].Advances in Neural Information Processing Systems,2001,13(6):556-562.
[5] LIN C J. Projected gradient methods for nonnegative matrix factorization[J]. Neural Computation,2007,19(10):2756-2779.
[6] ROBERTS D A,GARDNER M,CHURCH R,et al. Mapping chaparral in the santa monica mountains using multiple endmember spectral mixture models[J]. Remote Sensing of Environment,1998,65(3):267-279.
[7] ROESSNER S,SEGL K,HEIDEN U,et al. Automated differentiation of urban surfaces based on airborne hyperspectral imagery[J]. IEEE Transactions on Geoscience and Remote Sensing,2001,39(7):1525-1532.
[8] ROGGE D M,RIVARD B,ZHANG J K,et al. Iterative spectral unmixing for optimizing per-pixel endmember sets[J]. IEEE Transactions on Geoscience and Remote Sensing,2006,44(12):3725-3736.
[9] NASCIMENTO J M P,DIAS J M B. Vertex component analysis:a fast algorithm to unmix hyperspectral data[J]. IEEE Transactions on Geoscience and Remote Sensing,2005,43(4):898-910.
[10] 孔祥兵. 基于同质区分析的高光谱影像混合像元分解[D]. 武汉:武汉大学,2012:27-41.
[11] 孔祥兵,舒宁,陶建斌,等. 一种基于多特征融合的新型光谱相似性测度[J]. 光谱学与光谱分析,2011,31(8):2166-2170.
[12] HEINZ D C,C I C. Fully constrained least squares linear spectral mixture analysis method for material quantification in hyperspectral imagery[J]. IEEE Transactions on Geoscience and Remote Sensing,2001,39(3):529-545.
[13] LEE D D,SEUNG H S. Learning the parts of objects by non-negative matrix factorization[J]. Nature,1999,401(6755):788-791.
[14] KRUSE F A,LEFKOFF A B,BOARDMAN J W,et al. The spectral image processing system(SIPS)—interactive visualization and analysis of imaging spectrometer data[J]. Remote Sensing of Environment,1993,44(2/3):145-163.