Abstract
Spectral unmixing is an important technique for quantitatively analyzing hyperspectral remote sensing images. Recently, constrained nonnegative matrix factorization (NMF) has been demonstrated to be a powerful tool for spectral unmixing. However, acquiring the problem-dependent prior knowledge and incorporating it into NMF as effective constraints is a challenging task. In this article, a multiple clustering guided NMF unmixing approach is proposed under a self-supervised framework, which has been used to effectively learn high-level semantic information from the data with a surrogate task in many applications. Specifically, in order to provide self-supervised information to guide the NMF-based unmixing model, multiple clustering is integrated into the optimization process of NMF. Moreover, by introducing interaction between each clustering and the unmixing procedure, more accurate proximate endmember signatures and proximate abundance distributions are expected to be acquired and used to impose self-supervised constraints on endmembers and abundances, respectively. Consequently, effective prior information about endmember signatures and abundance distributions is captured and simultaneously integrated into NMF as valuable constraints to promote unmixing performance. Experiments are conducted on both synthetic data and real hyperspectral images, and the superior performance of our method is shown by comparing it with several state-of-the-art algorithms.
Highlights
H YPERSPECTRAL images (HSIs) are usually captured using many different electromagnetic bands and can contain rich spatial and spectral information about the observed scene, and thereby have many real applications
minimum volume constrained NMF (MVCNMF) utilizes the geometrical property of HSIs to improve the unmixing performance of nonnegative matrix factorization (NMF), and it attempts to minimize the volume of the simplex enclosed by the candidate endmembers in the unmixing procedure
When comparing all the algorithms based on the average values w.r.t. all the endmembers, our method shows the best performance by providing the lowest spectral angle distance (SAD)
Summary
H YPERSPECTRAL images (HSIs) are usually captured using many different electromagnetic bands and can contain rich spatial and spectral information about the observed scene, and thereby have many real applications. Limited by the low spatial resolution of the sensors, each pixel of an HSI usually cover a relatively large ground area, so that its spectral information may be a mixture of several pure spectra (i.e., endmembers). These mixed pixels sometimes have serious implications for quantitative analysis of HSIs. One of the techniques to tackle this problem is hyperspectral unmixing (HU), by which a set of endmembers included in an HSI and the corresponding proportions (i.e., abundances) of them in each pixel are estimated. NLSMM is mainly adopted to express a complex spectral mixing mechanism caused by multiple scattered effects of source radiation among several endmembers. Owing to the flexibility and tractability, LSMM is a widely used model to express the mixture mechanism of HSIs
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
More From: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.