22 May 2014 Nonlinear hyperspectral unmixing based on constrained multiple kernel NMF
Author Affiliations +
Nonlinear spectral unmixing constitutes an important field of research for hyperspectral imagery. An unsupervised nonlinear spectral unmixing algorithm, namely multiple kernel constrained nonnegative matrix factorization (MKCNMF) is proposed by coupling multiple-kernel selection with kernel NMF. Additionally, a minimum endmemberwise distance constraint and an abundance smoothness constraint are introduced to alleviate the uniqueness problem of NMF in the algorithm. In the MKCNMF, two problems of optimizing matrices and selecting the proper kernel are jointly solved. The performance of the proposed unmixing algorithm is evaluated via experiments based on synthetic and real hyperspectral data sets. The experimental results demonstrate that the proposed method outperforms some existing unmixing algorithms in terms of spectral angle distance (SAD) and abundance fractions.
© (2014) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Jiantao Cui, Jiantao Cui, Xiaorun Li, Xiaorun Li, Liaoying Zhao, Liaoying Zhao, "Nonlinear hyperspectral unmixing based on constrained multiple kernel NMF", Proc. SPIE 9124, Satellite Data Compression, Communications, and Processing X, 91240N (22 May 2014); doi: 10.1117/12.2057580; https://doi.org/10.1117/12.2057580

Back to Top