Unsupervised Cluster-based Band Selection for Hyperspectral Image Classification
- DOI
- 10.2991/icacsei.2013.135How to use a DOI?
- Keywords
- Hyperspectral image, band selection, dimensionality reduction, Landsat.
- Abstract
A hyperspectral image usually has a large data volume. Several dimensionality reduction (DR) approaches have been investigated to remove redundant information from highly corrected bands. One of the DR approaches is the unsupervised cluster-based band selection (UCBS) method, that is, the bands can be grouped together using different cluster strategies. The method is time consuming, however, because of iterative processing in the band cluster-stage. In this paper, an unsupervised cluster-based band selection method is proposed. The method including two steps is called SensorClust. Firstly the cross-correlation matrix of the entire image was computed and Landsat ETM+ sensor wavelength ranges were used to cluster bands. Secondly in each cluster the covariance matrix was computed, and the bands were selected with the maximum or minimum values along the diagonal of covariance matrix. To demonstrate the effectiveness of the proposed method, a support vector machine (SVM) was selected to carry out supervised classification. The experimental results show the proposed method achieves good classification results in terms of robust clustering.
- Copyright
- © 2013, 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 - Jee Cheng Wu AU - Gwo Chyang Tsuei PY - 2013/08 DA - 2013/08 TI - Unsupervised Cluster-based Band Selection for Hyperspectral Image Classification BT - Proceedings of the 2013 International Conference on Advanced Computer Science and Electronics Information (ICACSEI 2013) PB - Atlantis Press SP - 562 EP - 565 SN - 1951-6851 UR - https://doi.org/10.2991/icacsei.2013.135 DO - 10.2991/icacsei.2013.135 ID - Wu2013/08 ER -