摘要
最佳波段选择是高光谱影像降维的常用手段,将本征维数估计与核偏最小二乘法,相结合,提出一种基于核偏最小二乘法的最佳波段选择方法。首先利用自适应最大似然法估计高光谱数据的本征维数;然后将核方法引入到偏最小二乘法中,利用核偏最小二乘法对高光谱影像进行最佳波段选择,所需选择的波段数即为本征维数。实验分析表明,与其他最佳波段选择方法比较,本文方法输出的最佳波段用于地物分类,取得了较高的分类精度。
Optimal band selection is a common means of dimensionality reduction of hyperspectral images. Combi- ning intrinsic dimension estimation with partial least squares, an optimal band selection method based on kernel partial least squares was proposed. Adaptive maximum likelihood method is used to estimate the intrinsic dimension of hyperspectral data, and then the kernel method is introduced into the partial least squares. The number of the optimal bands of hyperspectral images which are selected with kernel partial least squares is intrinsic dimension. The experimental results showed that, comparing with other optimal band selection methods, the higher classifica- tion accuracy has been got with the optimal bands achieved from this method.
出处
《测绘科学技术学报》
CSCD
北大核心
2013年第2期172-176,共5页
Journal of Geomatics Science and Technology
基金
国家自然科学基金项目(41201477)
关键词
高光谱影像
最佳波段选择
自适应最大似然法
偏最小二乘
核方法
hyperspectral imagery
optimal band selection
adaptive largest likelihood method
partial least squares
kernel method
