A Novel Endmember Extraction Method Based
on Manifold Dimensionality Reduction
Xiaoyan Tang
(&) and Shangzheng Liu
School of Electronics and Electrical Engineering, Nanyang Institute
of Technology, Nanyang 473004, China
547717964@qq.com
Abstract. Because of multiple reflection and scattering, the mixed pixels in
hyperspectral images are actually nonlinear spectral mixing. Traditional endmember extraction algorithm is based on linear spectral mixing model, so the
extraction accuracy is not high. Aiming at the nonlinear structure of hyperspectral images, a novel endmember extraction method for hyperspectral images
based on Euclidean distance and nonlinear dimensionality reduction is proposed. This method introduces Euclidean distance of image into the nonlinear
dimensionality reduction algorithm of local tangent space permutation to
remove redundant spatial information and spectral dimensional information in
hyperspectral data and then extracts the endmembers from the reduced data by
searching for the maximum volume of the simplex. Experiments on real
hyperspectral data show that the proposed method has a good effect on hyperspectral image endmember extraction, and its performance is better than that of
linear dimensionality reduction PCA and original LTSA algorithm.
Keywords: Hyperspectral image Á Nonlinear dimensionality reduction Á
Euclidean distance of image Á Endmember extraction
1 Introduction
Due to the limitation of spatial resolution of hyperspectral imaging spectrometer and
the complexity and diversity of natural objects, some pixels of remote sensing images
often contain many types of objects, namely mixed pixels [1]. How to effectively
decompose mixed pixels has become an important direction in remote sensing research.
Decomposition of mixed pixel includes two steps: endmember extraction and abundance estimation. Endmember extraction is the precondition of mixed pixel decomposition. At present, there are many endmember extraction algorithms, including PPI,
VCA, N-FINDR [2], IEA [3], MVT, MNF [4, 5], and so on. These algorithms are
based on the linear spectral mixing model. However, due to the nonlinearity of ground
scattering, multiple scattering in a pixel, and the heterogeneity of sub-pixel components, hyperspectral data has inherent nonlinearity [6, 7].
One of the strategies to deal with the nonlinearity of hyperspectral images is
nonlinear dimensionality reduction. In Ref. [4], isometric mapping (ISOMAP), a
classical manifold learning algorithm, is introduced into hyperspectral image processing to replace PCA or MNF to reduce the dimensionality nonlinearly of
© Springer Nature Singapore Pte Ltd. 2020
Q. Liang et al. (Eds.): Artificial Intelligence in China, LNEE 572, pp. 615–621, 2020.
https://doi.org/10.1007/978-981-15-0187-6_73
on Manifold Dimensionality Reduction
Xiaoyan Tang
(&) and Shangzheng Liu
School of Electronics and Electrical Engineering, Nanyang Institute
of Technology, Nanyang 473004, China
547717964@qq.com
Abstract. Because of multiple reflection and scattering, the mixed pixels in
hyperspectral images are actually nonlinear spectral mixing. Traditional endmember extraction algorithm is based on linear spectral mixing model, so the
extraction accuracy is not high. Aiming at the nonlinear structure of hyperspectral images, a novel endmember extraction method for hyperspectral images
based on Euclidean distance and nonlinear dimensionality reduction is proposed. This method introduces Euclidean distance of image into the nonlinear
dimensionality reduction algorithm of local tangent space permutation to
remove redundant spatial information and spectral dimensional information in
hyperspectral data and then extracts the endmembers from the reduced data by
searching for the maximum volume of the simplex. Experiments on real
hyperspectral data show that the proposed method has a good effect on hyperspectral image endmember extraction, and its performance is better than that of
linear dimensionality reduction PCA and original LTSA algorithm.
Keywords: Hyperspectral image Á Nonlinear dimensionality reduction Á
Euclidean distance of image Á Endmember extraction
1 Introduction
Due to the limitation of spatial resolution of hyperspectral imaging spectrometer and
the complexity and diversity of natural objects, some pixels of remote sensing images
often contain many types of objects, namely mixed pixels [1]. How to effectively
decompose mixed pixels has become an important direction in remote sensing research.
Decomposition of mixed pixel includes two steps: endmember extraction and abundance estimation. Endmember extraction is the precondition of mixed pixel decomposition. At present, there are many endmember extraction algorithms, including PPI,
VCA, N-FINDR [2], IEA [3], MVT, MNF [4, 5], and so on. These algorithms are
based on the linear spectral mixing model. However, due to the nonlinearity of ground
scattering, multiple scattering in a pixel, and the heterogeneity of sub-pixel components, hyperspectral data has inherent nonlinearity [6, 7].
One of the strategies to deal with the nonlinearity of hyperspectral images is
nonlinear dimensionality reduction. In Ref. [4], isometric mapping (ISOMAP), a
classical manifold learning algorithm, is introduced into hyperspectral image processing to replace PCA or MNF to reduce the dimensionality nonlinearly of
© Springer Nature Singapore Pte Ltd. 2020
Q. Liang et al. (Eds.): Artificial Intelligence in China, LNEE 572, pp. 615–621, 2020.
https://doi.org/10.1007/978-981-15-0187-6_73