基于权重变分模型的地基IDOAS条带噪声去除  

Ground-Based IDOAS De-Striping by Weighted Unidirectional Variation

作　　者：奚亮 司福祺 江宇 周海金 邱晓晗 常振 XI Liang;SI Fu-qi;JIANG Yu;ZHOU Hai-jin;QIU Xiao-han;CHANG Zhen(Anhui Institute of Optics and Fine Mechanics,Hefei Institutes of Physical Science,Chinese Academy of Sciences,Hefei 230031,China;University of Science and Technology of China,Hefei 230026,China)

机构地区：[1]中国科学院合肥物质科学研究院,安徽光学精密机械研究所,安徽合肥230031 [2]中国科学技术大学,安徽合肥230026

出　　处：《光谱学与光谱分析》2022年第2期627-633,共7页Spectroscopy and Spectral Analysis

基　　金：国家重点研发计划项目(2018YFC0213201,2019YFC0214702)资助。

摘　　要：成像差分吸收光谱技术是成像光谱技术和差分吸收光谱技术的结合,能够采集图谱合一的数据立方,并通过光谱反演得到痕量气体浓度的二维分布信息。地基IDOAS仪器通过安装平台的水平旋转实现摆扫成像,可用于识别污染气体的排放源和监测气体的扩散情况。然而和所有的成像光谱技术相类似,地基IDOAS也容易出现条带噪声的问题,会产生相应的伪结构,影响后续的信息提取和数据分析。目前星载和机载IDOAS中常见的条带噪声去除算法有均匀区域校正法、传输模型模拟法、傅里叶变换频域滤波法、多项式拟合法等,应用到地基仪器中均存在不适用的问题。介绍了一种基于权重变分模型的条带噪声去除算法,该算法首先通过分块自适应阈值分割得出表征遮挡区域的权重矩阵,然后利用条带噪声的方向性和稀疏性建立各向异性的变分模型,最后通过交替方向乘子算法迭代求解。为检验去条带算法的可靠性,使用稀疏、稠密、周期、随机、整行、部分、单行、多行等多种模拟噪声进行了性能测试。测试结果证明权重变分算法能够有效去除各种常见的条带噪声,目视效果和四种全参考评价指标均有良好的表现。地基IDOAS于2018年夏季在四川乐山进行了外场实验,实验中仪器的水平扫描范围覆盖360°全方位角,扫描间隔为1°,垂直方向仪器同时采集0°~30°仰角内的光谱。仪器的积分时间设置为500 ms,每组全景扫描的工作时间约为15 min。利用DOAS技术对采集到的太阳散射光谱进行反演,最终得到的NO_(2)和SO_(2)气体的二维浓度分布图的像素大小为360×48。从反演结果来看,条带噪声对不同时间和不同气体的观测结果的影响大小均不同。经权重变分算法处理后,多组NO_(2)和SO_(2)浓度分布中的条带噪声情况得到极大的改善,并且没有出现过度平滑的情况。结果表明,该算法适用于�Imaging differential optical absorption spectroscopy technology(IDOAS)combines imaging spectroscopy and differential optical absorption spectroscopy.The acquired data of IDOAS instruments are so-called hyperspectral cube with two spatial dimensions and a spectral one.After DOAS analysis of the original data,two-dimensional trace gas distributions can be resolved.For ground-based IDOAS instruments,the imaging capability is achieved through the stepwise rotation of the motor in the horizontal direction,which can be used to identify the emission sources of polluting gases and monitor the transmission of pollution.However,similar to other imaging spectroscopy instruments,ground-based IDOAS instruments are also prone to stripe noise,producing corresponding pseudo-structures and affecting subsequent information extraction and data analysis.Several de-striping algorithms have been applied for space-borne and airborne sensors,including homogeneous reference area correction method,transmission model simulation method,frequency domain filtering method,polynomial fitting method,which are not fully applicable to ground-based instruments.Here we present a de-striping algorithm based on a weighted unidirectional variation model.This algorithm first obtains a weight matrix that characterizes the blocked area through adaptive threshold segmentation,then utilizes the unidirectionality and sparsity of the strip noise to establish the optimization model,which is solved iteratively by using the alternating direction method multipliers technique.To test the performance of the de-striping algorithm,simulated experiments were performed using various cases including sparse,dense,periodic,random,whole-line,partial,single-line,multi-line stripe noise.Corresponding results prove that this algorithm can effectively remove typical stripe noise,with good performances in visual and four full-reference evaluations.Ground-based IDOAS observations were carried out in Leshan,Sichuan province,in the summer of 2018.In this experiment,the IDOAS instr

关 键 词：成像差分吸收光谱 条带噪声 变分法 光学遥感 

分 类 号：O433[机械工程—光学工程]