Journal
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
Volume 2, Issue 2, Pages 156-159Publisher
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/LGRS.2005.844658
Keywords
band selection; feature weighting; per-class decorrelation; hyperspectral data classification
Ask authors/readers for more resources
A new feature weighting method for band selection is presented, which is based on the pairwise separability criterion and matrix coefficients analysis. Through decorrelation of each class by principal component transformation, the criterion value of any band subset is the summations of the values of individual bands of it for the transformed feature space, and thus the computation amounts of calculating criteria of each band combinations are reduced. Following it, the corresponding matrix coefficients analysis is done to assign weights to original bands. As feature weighting considers little about the spectral correlation, the redundant bands are removed by choosing those with lower correlation coefficients than a preset threshold. Hyperspectral data classification experiments show the effectiveness of the new band selection method.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available