Translator Disclaimer
2 February 2009 Endmember extraction by pure pixel index algorithm from hyperspectral image
Author Affiliations +
We describe and validate an automated methodology based on PPI to extract endmembers from images and distinct the according endmembers. Four main steps are:1)project the raw image cube to its most spectral dimensions and non-noise components by minimum noise fraction (MNF) technology; 2) use the set of spectrally distinct pixels produced by MNF as skewers for PPI, generates a list of candidates from which final endmembers can be selected; 3) an automatic selection procedure based on K-means clustering is consequently performed to determined the centriod of endmenbers. 4) linear spectral mixing model (LSMM) is used to estimate mixing coefficient. And root mean square error (RMSE) reflects the accuracy of decomposition. We use the methodology to investigate the unique properties of hyperspectral data and how spectral information can be used to identify mineralogy with the Airborne Visible/infrared imaging Spectrometer (AVIRIS) hyperspectral data from Cuprite, Nevada.
© (2009) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Wenyu Wang and Guoyin Cai "Endmember extraction by pure pixel index algorithm from hyperspectral image", Proc. SPIE 7157, 2008 International Conference on Optical Instruments and Technology: Advanced Sensor Technologies and Applications, 71570E (2 February 2009);

Back to Top