Abstract

Interpretation of mineral hyperspectral images provides large amounts of high-dimensional data, which is often complicated by mixed pixels. The quantitative interpretation of hyperspectral images is known to be extremely difficult when three types of information are unknown, namely, the number of pure pixels, the spectrum of pure pixels, and the mixing matrix. The problem is made even more complex by the disturbance of noise. The key to interpreting abstract mineral component information, i.e., pixel unmixing and abundance inversion, is how to effectively reduce noise, dimension, and redundancy. A three-step procedure is developed in this study for quantitative interpretation of hyperspectral images. First, the principal component analysis (PCA) method can be used to process the pixel spectrum matrix and keep characteristic vectors with larger eigenvalues. This can effectively reduce the noise and redundancy, which facilitates the abstraction of major component information. Second, the independent component analysis (ICA) method can be used to identify and unmix the pixels based on the linear mixed model. Third, the pure-pixel spectrums can be normalized for abundance inversion, which gives the abundance of each pure pixel. In numerical experiments, both simulation data and actual data were used to demonstrate the performance of our three-step procedure. Under simulation data, the results of our procedure were compared with theoretical values. Under the actual data measured from core hyperspectral images, the results obtained through our algorithm are compared with those of similar software (Mineral Spectral Analysis 1.0, Nanjing Institute of Geology and Mineral Resources). The comparisons show that our method is effective and can provide reference for quantitative interpretation of hyperspectral images.

End-member extraction based on segmented vertex component analysis in hyperspectral images

Mingyu Nie, Zhi Liu, Xiaofu He, Qingchen Qiu, Yuanyuan Zhang, and Ju Chang
Appl. Opt. 56(9) 2476-2482 (2017)

Using independent component analysis for material estimation in hyperspectral images

Chia-Yun Kuan and Glenn Healey
J. Opt. Soc. Am. A 21(6) 1026-1034 (2004)

Structured covariance principal component analysis for real-time onsite feature extraction and dimensionality reduction in hyperspectral imaging

Jaime Zabalza, Jinchang Ren, Jie Ren, Zhe Liu, and Stephen Marshall
Appl. Opt. 53(20) 4440-4449 (2014)

Graph-regularized tensor robust principal component analysis for hyperspectral image denoising

Yongming Nie, Linsen Chen, Hao Zhu, Sidan Du, Tao Yue, and Xun Cao
Appl. Opt. 56(22) 6094-6102 (2017)

A quantitative spectral component analysis method based on maximum likelihood

Xiu-Zuo Liang, Xiao-Yu Pang, Zhi-Ming Zhang, Long Wei, and Lei Shuai
Opt. Express 27(5) 7160-7173 (2019)