Hyperspectral unmixing is a popular research topic in hyperspectral processing, aiming at obtaining the ground features contained in the mixed pixels and their proportion. Recently, nonlinear mixing models have received particular attention in hyperspectral decomposition since the linear mixing model cannot suitably apply in the situation that exists in multiple scattering. In this study, we constructed a residual dense autoencoder network (RDAE) for nonlinear hyperspectral unmixing in multiple scattering scenarios. First, an encoder was built based on the residual dense network (RDN) and attention layer. The RDN is employed to characterize multi-scale representations, which are further transformed with the attention layer to estimate the abundance maps. Second, we designed a decoder based on the unfolding of a generalized bilinear model to extract endmembers and estimate their second-order scattering interactions. Comparative experiments between the RDAE and six other state-of-the-art methods under synthetic and real hyperspectral datasets demonstrate that the proposed method achieved a better performance in terms of endmember extraction and abundance estimation.
With the development of hyperspectral technology, it has become possible to classify alteration zones using hyperspectral data. Since various altered rocks are comprehensive manifestations of mineral assemblages, their spectra are highly similar, which greatly increases the difficulty of distinguishing among them. In this study, a Semi-Supervised Adversarial Autoencoder (SSAAE) was proposed to classify the alteration zones, using the drill core hyperspectral data collected from the Pulang porphyry copper deposit. The multiscale feature extractor was first integrated into the encoder to fully exploit and mine the latent feature representations of hyperspectral data, which were further transformed into discrete class vectors using a classifier. Second, the decoder reconstructed the original inputs with the latent and class vectors. Third, we imposed a categorical distribution on the discrete class vectors represented in the one-hot form using the adversarial regularization process and incorporated the supervised classification process into the network to better guide the network training using the limited labeled data. The comparison experiments on the synthetic dataset and measured hyperspectral dataset were conducted to quantitatively and qualitatively certify the effect of the proposed method. The results show that the SSAAE outperformed six other methods for classifying alteration zones. Moreover, we further displayed the delineated results of the SSAAE on the cross-section, in which the alteration zones were sensible from a geological point of view and had good spatial consistency with the occurrence of Cu, which further demonstrates that the SSAAE had good applicability for the classification of alteration zones.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.