一种在线字典学习的高光谱图像盲解混方法
Blind Unmixing of Hyperspectral Images Based on Online Dictionary Learning
宋晓瑞 1赵忠文 2于尧1
作者信息
- 1. 装备学院 研究生管理大队, 北京 101416
- 2. 装备学院 复杂电子系统仿真实验室, 北京 101416
- 折叠
摘要
针对待处理高光谱图像中所含地物光谱未知的情况,将在线字典学习的方法引入高光谱稀疏解混中,提出了一种基于在线字典学习和稀疏编码的高光谱图像盲解混方法.通过在线字典学习以及对稀疏编码的统计,从训练字典中筛选出与未知地物光谱最为接近的原子,作为待处理数据所含地物光谱的估计.仿真结果表明:这种方法的准确提取概率超过66%,有效提取概率超过89%,对地物光谱识别等相关研究有一定借鉴意义;另外,利用训练字典和稀疏编码可较好的重构混合像素光谱向量.
Abstract
In view of the unknown condition of the surface feature spectra in the hyperspectral images to be processed, with the introduction of an online dictionary learning into the hyperspectral sparse unmixing, this paper proposes a method of blind hyperspectral unmixing based on online dictionary learning and sparse coding.Atoms which are closest to the unknown surface feature spectra are selected from the training dictionary via the online dictionary learning and sparse coding statistics.They are used to estimate the surface feature spectra in the data to be processed.The simulation results show that the accurate extraction probability of this method is more than 66%, and the effective extraction probability is more than 89%.It can be used for reference in the research of spectral identification of ground objects.In addition, the training dictionary and sparse coding can be used to reconstruct the mixed pixel spectral vectors.Therefore, this method also works well in sparse unmixing.
关键词
高光谱遥感/在线字典学习/稀疏编码/地物光谱提取/稀疏解混Key words
hyperspectral remote sensing/online dictionary learning/sparse coding/surface feature spectrum extraction/sparse unmixing引用本文复制引用
出版年
2017
NETL
NSTL
维普
万方数据