An Informed ADMM-Based Approach for Hyperspectral Unmixing Addressing Additively-Adapted Spectral Variability
Yasmine Kheira Benkouider, Fatima Zohra Benhalouche, Meziane Iftene, Moussa Sofiane Karoui
Abstract
Yasmine Kheira Benkouider, Fatima Zohra Benhalouche, Meziane Iftene, Moussa Sofiane Karoui
Abstract
In remote sensing, hyperspectral unmixing aims to identify the pure spectral signatures, known as "endmembers", present within each pixel of a hyperspectral image and to estimate the abundance fractions of each endmember. Hyperspectral unmixing accuracy is significantly affected by the spectral variability phenomenon. In this paper, a new informed hyperspectral unmixing approach, which deals with additively-adapted spectral variability, is introduced. This approach, in addition to taking into account the spectral variability phenomenon, exploits one or more known pure material spectra present in considered hyperspectral data. Based on the alternating direction multiplier method, the proposed method decomposes the observed mixed pixel spectra into constituent spectral signatures and corresponding abundance fractions. Simulation results, using semi-real data, demonstrate the robustness of the proposed method for unmixing hyperspectral data with spectral variability.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
In remote sensing, hyperspectral unmixing aims to identify the pure spectral signatures, known as "endmembers", present within each pixel of a hyperspectral image and to estimate the abundance fractions of each endmember. Hyperspectral unmixing accuracy is significantly affected by the spectral variability phenomenon. In this paper, a new informed hyperspectral unmixing approach, which deals with additively-adapted spectral variability, is introduced. This approach, in addition to taking into account the spectral variability phenomenon, exploits one or more known pure material spectra present in considered hyperspectral data. Based on the alternating direction multiplier method, the proposed method decomposes the observed mixed pixel spectra into constituent spectral signatures and corresponding abundance fractions. Simulation results, using semi-real data, demonstrate the robustness of the proposed method for unmixing hyperspectral data with spectral variability.
Key concepts: Hyperspectral imaging, Endmember, Robustness (evolution), Pixel, Full spectral imaging, Spectral signature, Remote sensing, Computer science