摘要
基于凸面几何学理论,由端元作为角点的单形体的体积应该是最大的。著名的N-FINDR和SGA算法正是基于以上理论,通过在数据云中寻找体积最大的单形体来实现端元的自动提取。本文利用粒子群优化(PSO)技术,基于凸面几何学理论,设计了一个新的端元提取算法。利用模拟和真实高光谱影像对其进行了实验,并将其结果与N-FINDR和SGA算法的结果进行了比较分析。
Based on convex geometry theory, the simplex with vertices that are given by the spectra of the endmembers has the big- gest volume. N-FINDR and SGA which based on convex geometry theory are popular endmember extraction algorithms. They select the simplex which has the biggest volume in the data cloud to extracting endmembers automatically. An endmember extraction method that is based on Particle Swarm Optimization （PSO） and convex geometry theory was developed in this paper. It carried out the experiments by simulative and real hyperspectral imageries. As well as it compared and analyzed the results among the PSO, N-FINDR and SGA.
出处
《测绘科学》
CSCD
北大核心
2011年第4期16-18,30,共4页
Science of Surveying and Mapping
关键词
高光谱影像
粒子群算法
线性混合模型
端元提取
N—FINDR
hyperspectral imagery
particle swarm optimization
linear mixture model
endmember extraction
N-FINDR
