摘要
基于稀疏的高光谱解混方法作为一类流行的解混方法可以获得较为理想的解混结果。但现有字典裁剪方法只通过一种字典裁剪来得到光谱库子集,会导致得到的光谱库子集不够准确。为提高解混的精度,提出将光谱信息散度和光谱角制图相减作为两次字典裁剪方法(Spectral Information Divergence minus Spectral Angle Mapping,SS)。两次字典裁剪较一重字典裁剪进一步降低光谱特征不匹配对解混精度的影响,可改善稀疏解混的性能。该文将提出的SS与光谱信息散度、光谱角制图、鲁棒的多重信号分类4种字典裁剪方法用在联合稀疏块低秩解混算法中以来证明两次字典裁剪方法的有效性。
As a popular unmixing method,Hyperspectral unmixing method based on sparsity can obtain ideal unmixing results.However,the existing dictionary clipping methods only use one dictionary clipping to obtain the spectrum library subset,which will lead to the inaccurate spectrum library subset.In order to improve the accuracy of unmixing,this paper proposes aSpectral Information Divergence minus Spectral Angle Mapping(SS)method for twice dictionary clipping method.Twice dictionary clipping can further reduce the influence of spectral feature mismatch on the unmixing accuracy and improve the performance of sparse unmixing.In this paper,the proposed four dictionary clipping methods of SS and spectral information divergence and spectral angle mapping and robust multiple signal classification are used in the joint sparse block low rank unmixing algorithm to prove the effectiveness of the twice dictionary clipping method.
作者
张子龙
沈珣
阎庚未
ZHANG Zilong;SHEN Xun;YAN Gengwei
出处
《科技创新与应用》
2023年第29期22-25,30,共5页
Technology Innovation and Application
关键词
高光谱图像
高光谱解混
稀疏解混
字典裁剪
光谱库子集
hyperspectral image
hyperspectral unmixing
sparse unmixing
dictionary clipping
spectral library subset
