摘要
提出一种基于低秩约束惩罚最小二乘(LRPLS)的基线校正方法。利用惩罚最小二乘模型综合考虑拟合基线对干涉图的保真度和平滑度的影响,同时引入有效干涉图和噪声的低秩-稀疏先验约束条件,从而构建联合低秩矩阵恢复和惩罚最小二乘的正则化框架,并采用基于增广拉格朗日乘子的迭代优化算法进行求解。在“嫦娥一号”卫星干涉成像光谱仪(IIM)数据上的实验表明,所提方法在去除基线的同时,能够保留干涉图的有效信息。相比于现有的基线校正方法,所提方法具有更好的稳定性和抗噪能力;校正后复原高光谱影像得到显著提升,对于改善干涉成像光谱仪数据质量具有较高的实际应用价值。
A baseline correction method based on low-rank constraint and penalized least squares(LRPLS)is proposed.This paper comprehensively considers the fitted baseline′s fidelity to the interferogram and its own smoothness by using the penalized least squares model.At the same time,the low-rank and sparse prior constraint conditions of the effective interferogram and noise are introduced to build a regularization framework combing the low-rank matrix recovery and penalized least squares,and the solution is carried out by an augmented Lagrangian multiplier based iterative optimization algorithm.The experiments on the data of Chang′e-1 interference imaging spectrometer(IIM)show that the proposed method can retain effective information of the interferogram while removing the baseline.Compared with existing baseline correction methods,the proposed method has better stability and anti-noise ability.In addition,the recovered hyperspectral image significantly improves after baseline correction,which has high practical significance for improving the data quality of IIM.
作者
朱锋
安军社
施海亮
叶函函
李志伟
王先华
熊伟
Zhu Feng;An Junshe;Shi Hailiang;Ye Hanhan;Li Zhiwei;Wang Xianhua;Xiong Wei(Key Laboratory of Electronics and Information Technology for Spacing Systems,National Space Science Center,Chinese Academy of Sciences,Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China;Anhui Institute of Optics and Fine Mechanics,Hefei Institute of Physical Science,Chinese Academy of Sciences,Hefei 230031,Anhui,China)
出处
《光学学报》
EI
CAS
CSCD
北大核心
2022年第14期216-222,共7页
Acta Optica Sinica
基金
国家重点研发计划(2016YFB0500704)
中科院重点部署项目(ZDRW-KT-2020-3)。
关键词
光谱学
干涉成像光谱仪
基线校正
光谱复原
惩罚最小二乘
低秩矩阵恢复
spectroscopy
interference imaging spectrometer
baseline correction
spectrum recovery
penalized least squares
low-rank matrix recovery
