摘要
高光谱稀疏解混合方法旨在寻找光谱库的最佳子集以对场景中的混合像元进行建模。大多数稀疏解混合方法使用美国地质调查局光谱库,易造成与所研究的高光谱数据失配。利用顶点分量分析和概率输出支持向量机,设计了一种结合空间和光谱信息的基于图像的端元光谱库提取方法。由于提取的端元光谱库具有群结构,即多个端元光谱代表一类材料,因此估计的丰度也具有群体结构。提出了基于分数群稀疏混合范式和空间正则化的解混算法,用来解决丰度估计优化问题。分数混合范式诱导丰度群内和群间稀疏性,全变分(Total variation,TV)空间正则化诱导丰度群空间平滑。在模拟数据集和真实数据集上的实验结果表明,与传统稀疏解混合方法相比,该方法可以显著提高解混性能。
Hyperspectral sparse unmixing methods aim at finding an optimal subset of a spectral library that can best model a mixed pixel in the scene.Most sparse unmixing methods utilize the United States Geological Survey spectral library which may not match the given hyperspectral data.Using vertex component analysis and probabilistic outputs for support vector machines,the image-based spectral library extraction method incorporating spatial and spectral information is designed.In the image-based spectral library,one material is represented with a group of spectra.Consequently,the estimated abundances also possess a group structure.A fractional mixed norm and a total variation spatial regularization are proposed to solve the abundance estimation optimization problem.The fractional mixed norm enforces group and within-group sparsity,and the Total variation(TV)spatial regularization enforces group smoothing.The results of experiments on simulated and real datasets reveal that the proposed method can significantly improve the unmixing performance compared with the classical sparse unmixing methods.
作者
王伞
王立国
王丽凤
WANG San;WANG Liguo;WANG Lifeng(College of Information and Communications Engineering,Harbin Engineering University,Harbin 150001,China)
出处
《黑龙江大学自然科学学报》
CAS
2023年第3期332-340,共9页
Journal of Natural Science of Heilongjiang University
基金
国家自然科学基金(62071084)。
关键词
稀疏解混
基于图像的光谱库提取
群稀疏混合范式
全变分空间正则
sparse unmixing
image-based spectral library extraction
group sparsity mixed norm
total variation spatial regularization
