选剔同步的高光谱遥感图像波段选择算法

为了降低高光谱遥感图像冗余度,减少后续的计算复杂度,提出了选剔同步的高光谱遥感图像波段选择算法。以主成分分析后的数据作为参考波段来源,以互信息作为选取波段的相似性度量,引入R-KL系数作为剔除波段的判别准则,利用边选取边剔除的方式进行波段选择。为了验证该算法的有效性,运用贝叶斯分类法对降维后波段进行分类,并与自适应波段选择和基于最大信息量的波段选择算法进行比较。结果显示当选取波段数目较少时,该算法的分类效果优于上述两种算法,当选取波段数目较多时,3种算法分类效果相当,故该算法是一种有效的波段选择算法。

Band-Selection Algorithm for Hyperspectral Imagery with Simultaneous Selection and Elimination

To reduce the redundancy of hyperspectral remote-sensing images and the subsequent computational complexity,a band-selection algorithm with simultaneous selection and elimination is proposed.Using both the data derived from principal component analysis as a source for the reference band as well as related information on the similarity measure of the selected band,the R-KL ratio is introduced as the discriminant criterion to eliminate bands.The basic objective of the proposed algorithm is simultaneous band selection and rejection.To validate the proposed algorithm,Bayesian classification is used to classify the dimensionally reduced bands.The proposed algorithm is compared with the adaptive band selection method and the band-selection algorithm based on maximum information.When the number of selected bands is small,the results show that the classification effect of the proposed algorithm is much greater than the aforementioned two algorithms.When the number of selected bands is large,the three algorithms perform quite effectively.We can conclude that the proposed band-selection algorithm is effective at band selection and rejection.