摘要
根据核熵成分分析(KECA)的特点提出了基于凸面几何学概念的样本集选取方法和以特征空间光谱角为相似性度量的C-均值分类算法,并将其用于高光谱遥感图像分类。在HYDICE高光谱数据上的试验表明,本文提出的算法可以有效地提高分类精度。
To take advantage of the characteristics of KECA for hyperspectral remote sensing image classification,an approach of sample set selection and C-means classification is proposed.The sample selection is based on convex geometry concepts and C-means classification uses spectral angles as distance metrics in feature space.Experiment results of HYDICE hyperspectral data confirm that the proposed approach can improve classification accuracy effectively.
出处
《吉林大学学报(工学版)》
EI
CAS
CSCD
北大核心
2012年第6期1597-1601,共5页
Journal of Jilin University:Engineering and Technology Edition
基金
国家自然科学基金项目(60802084)
关键词
信息处理技术
高光谱图像
RENYI熵
图像分类
核熵成分分析
information processing
hyperspectral image
Renyi entropy
image classification
kernel entropy component analysis