The endmembers selection and spectral unmixing based on the optimal combination of the endmembers extracted by N-FINDR algorithm and SSWA algorithm
- DOI
- 10.2991/meic-14.2014.208How to use a DOI?
- Keywords
- hyperspectral; endmember; simplex; spectral unmixing
- Abstract
Based on the convex geometry, a lot of endmember extraction algorithms have been proposed, in which N-FINDR algorithm use sample points in data cloud to construct simplex and maximize its volumn, it suitable for the images which contain “pure pixels”, we can randomly select pixels as the endmembers to form the simplex and calculate its volume, and then use other pixels to replace one of the endmembers, calculate the simplex volume again. If the replacement volume increase, then the replacement is accepted, otherwise the replacement is given up, until all the endmembers are founded. SSWA algorithm constructs a simplex to surround the data cloud and shrink it continuously, that is to say, the extracted endmembers may have no corresponding "pure pixels" in the original hyperspectral images, it firstly finds a simplex to contain all the sample points, and then shrinks the simplex volume according to the gradient descent rule with a penalty function. This paper combines the advantages of these two algorithms, selects the optimal combination within the endmembers extracted by these two algorithms, finally obtains the best spectral unmixing accuracy.
- Copyright
- © 2014, the Authors. Published by Atlantis Press.
- Open Access
- This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).
Cite this article
TY - CONF AU - Jun Xu AU - Fuhong Xu PY - 2014/11 DA - 2014/11 TI - The endmembers selection and spectral unmixing based on the optimal combination of the endmembers extracted by N-FINDR algorithm and SSWA algorithm BT - Proceedings of the 2014 International Conference on Mechatronics, Electronic, Industrial and Control Engineering PB - Atlantis Press SP - 931 EP - 935 SN - 2352-5401 UR - https://doi.org/10.2991/meic-14.2014.208 DO - 10.2991/meic-14.2014.208 ID - Xu2014/11 ER -