光谱信息扩展下的纸张粘度光谱指数构建研究  

Study on the Derivation of Paper Viscosity Spectral Index Based on Spectral Information Expansion

作　　者：高宇 孙雪剑 李广华[2] 张立福 曲亮[2] 张东辉 常晶晶 戴晓爱[3] GAO Yu;SUN Xue-jian;LI Guang-hua;ZHANG Li-fu;QU Liang;ZHANG Dong-hui;CHANG Jing-jing;DAI Xiao-ai(National Engineering Laboratory for Remote Sensing Satellite Applications,Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100101,China;The Palace Museum,Beijing 100009,China;College of Earth Sciences,Chengdu University of Technology,Chengdu 610059,China)

机构地区：[1]中国科学院空天信息创新研究院,遥感卫星应用国家工程实验室,北京100101 [2]故宫博物院,北京100009 [3]成都理工大学地球科学学院,四川成都610059

出　　处：《光谱学与光谱分析》2023年第9期2960-2966,共7页Spectroscopy and Spectral Analysis

基　　金：国家重点研发计划项目(2019YFC1520405),国家重点研发计划项目(2020YFE0204600)资助。

摘　　要：粘度是反映纸张纤维素分子聚合度和力学性能的重要指标,实时、准确地获取粘度含量对于珍贵纸质材料的修复和保护具有重要的意义。然而,传统的纸张粘度分析方法主要采用化学手段,该化验过程耗时较长,且会对纸张不可避免的产生二次损伤。针对这一问题,高光谱遥感凭借其丰富信息量、实时、无接触的特点,是无损获取纸张粘度含量的有效途径。首先,在实验室获得不同老化程度的实验纸张测得其粘度含量,采集纸张样本的光谱影像数据,通过光谱降噪、光谱变换和光谱信息扩展实现纸张高光谱数据的预处理,建立不同老化程度下的纸张粘度含量光谱数据库,分别构建不同光谱变换方式下的光谱差值、比值和归一化指数,再结合相关性分析筛选其中与粘度相关性最强的12种最优光谱指数,最后将其作为自变量搭建关于粘度含量的回归模型,通过模型精度对比来优选其中最能有效表征纸张粘度含量变化的光谱指数及模型。研究结果表明:(1)相对于原始光谱而言,经过光谱变换处理后提取的粘度高相关特征子集占比得到大幅提升,同时其中的相关系数均值与中位数也得到提高;(2)通过光谱信息扩展后得到的光谱信息参量与粘度的相关性高于原本的光谱谱段,且提取的12种最优光谱指数中大部分有扩展信息参量的参与;(3)不同光谱变换结果下提取的最佳光谱指数与粘度含量的相关性都在0.89以上,由其中筛选得到的三种具备代表性的光谱指数都有效反映纸张粘度在400~500 mL·g^(-1)时的变化情况;(4)纸张光谱经过对数一阶微分处理后,由光谱积分(SI)和光谱吸收深度(SAD)构建的归一化指数与粘度相关性最大,达到了-0.917,由该指数建立的模型在训练集和测试集上R^(2)分别为0.84和0.76,其在测试集中MRE为0.089,RMSE为40.29 mL·g^(-1)。研究结果可为纸张粘度含量遥感反演提供科�Viscosity is an important index reflecting paper cellulose's degree of polymerization and physical properties.Accurate real time information on viscosity is important for repairing and protecting precious paper materials.However,the traditional paper's viscosity analysis method mainly uses chemical means,which takes a long time and will inevitably cause secondary damage to the paper.To solve this problem,hyperspectral remote sensing,with its rich information and real-time,contactless characteristics,is an effective way to obtain the paper's viscosity content without damage.First,obtain experimental papers with different aging degrees in the laboratory to measure their viscosity contents,collect hyperspectral data of paper samples,preprocess paper samples hyperspectral data through spectral noise reduction,spectral transformation,and spectral information expansion,establish a spectral database of paper's viscosity contents under different aging degrees,and respectively build spectral difference index,ratio index and normalization index under different spectral transformation methods.Correlation analyses were carried out the 12 best spectral indices with the strongest correlation with viscosity were selected.Finally,the selected spectral indices were used as independent variables to build a regression model on the viscosity content.We also selected the spectral index and model that best characterized the change of the paper's viscosity content by comparing model accuracy.The results show that:(1)Compared with the original spectrum,the proportion of the highly correlated feature subset of viscosity extracted after spectral transformation processing greatly improved along with the mean and median of the correlation coefficient;(2)The correlation between the spectral information parameters(obtained by spectral information expansion)and viscosity is higher than that of the original spectral segment,and most of the 12 optimal spectral indices extracted have the participation in the expanded information parameters;(3)The

关 键 词：纸张 粘度 高光谱遥感 光谱指数 遥感建模 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]