基于可见光-近红外高光谱成像技术的陈皮陈化年份鉴别研究  

Identification of Xinhui Citri Reticulatae Pericarpium of Different Aging Years Based on Visible-Near Infrared Hyperspectral Imaging

作　　者：张悦 周骏辉[1] 王思曼 王游游 张云昊 赵帅 刘舒扬[2] 杨健[1] ZHANG Yue;ZHOU Jun-hui;WANG Si-man;WANG You-you;ZHANG Yun-hao;ZHAO Shuai;LIU Shu-yang;YANG Jian(State Key Laboratory for Quality Ensurance and Sustainable Use of Dao-di Herbs,National Resource Center for Chinese Materia Medica,China Academy of Chinese Medical Sciences,Beijing 100700,China;Tianjin Jinhang Institute of Technical Physics,Tianjin 300381,China;School of Traditional Chinese Medicine,Yunnan University of Chinese Medicine,Kunming 650500,China)

机构地区：[1]道地药材品质保障与资源持续利用全国重点实验室,中国中医科学院中药资源中心,北京100700 [2]天津津航技术物理研究所,天津300381 [3]云南中医药大学中药学院,云南昆明650500

出　　处：《光谱学与光谱分析》2023年第10期3286-3292,共7页Spectroscopy and Spectral Analysis

基　　金：中国中医科学院科技创新工程项目(CI2021A04005);财政部和农业农村部:国家现代农业产业技术体系项目(CARS-21);国家产业技术基础公共服务平台项目:中药全产业链质量技术服务平台(2022-230-221)资助。

摘　　要：高光谱成像技术(HSI)是基于非常多窄波段的影像数据技术,将成像技术与光谱技术相结合,获取高光谱分辨率的连续、窄波段的图像数据,因其快速、无损的特点,被广泛应用于食品、农产品、中药材等样品的快速鉴别。道地药材新会陈皮具有较高的市场价值,且陈化(贮藏)年份越久市场价格亦越高,市场人工鉴别准确率低、难度大。基于高光谱技术结合化学计量学方法,建立不同陈化年份新会陈皮的快速无损鉴别方法。采集5个陈化年份样品在可见-近红外波段(400~1000 nm)的高光谱信息,提取高光谱图像感兴趣区域(ROI)的平均光谱值作为样本原始光谱。经黑白校正后获得标准数据,通过多元散射校正(MSC)、一阶导数(D1)、二阶导数(D2)、SG平滑(SG)和标准正态变量变换(SNV)5种预处理方法对数据进行降噪处理后,结合偏最小二乘判别分析(PLS-DA)、随机森林(RF)和支持向量机(SVM)等方法建立分类鉴别模型,以预测结果的准确率作为评价指标筛选最佳模型,使用混淆矩阵(confusion matrix)评估模型分类性能。结果表明,外表皮数据以MSC结合PLS-DA方法为最优鉴别模型,预测集鉴别准确率达到97.59%;而内表皮数据则以原始数据结合PLS-DA方法为最优鉴别模型,预测集鉴别准确率亦达到97.59%。采用内表皮数据,进一步采用连续投影算法(SPA)选择19个特征波长建模,整体判别准确率仍可达90%以上。通过SPA方法提取的特征波长建模可以达到与全波长模型相似的识别效果,去除冗余变量可以大大降低模型的复杂性,减少模型的运算时间。该研究建立的高光谱技术结合化学计量学的方法可实现不同陈化年份新会陈皮样品的快速无损鉴别,为专属小型化仪器装备系统的开发提供了理论依据。Hyperspectral imaging(HSI)is an image data technology based on narrow bands.It combines imaging with spectral technology to obtain continuous and narrow-band image data with high spectral resolution.Hyperspectral imaging technology is widely used in rapidly identifying food,agricultural products,Chinese medicinal materials and other samples because of its rapid and non-destructive characteristics.Xinhui Citri Reticulatae Pericarpium has a high market value,and the market price of the sample is higher because of the longer storage age.At the same time,the accuracy of manual identification of the tangerine peel market is difficult.Based on hyperspectral imaging and chemometric,this study established a rapid and nondestructive identification method for Xinhui Citri Reticulatae Pericarpium of different aging years.Hyperspectral information of 5 aged samples in the vision-near-infrared band(400~1000 nm)is collected.The average spectral value of the Region of interest(ROI)of the hyperspectral image was extracted as the original sample spectrum.Standard data were obtained after black-and-white correction.After denoising the data by 5 pretreatment methods,including Multiplicative scatter correction(MSC),first Derivative(D1)and Second Derivative(D2),SG smoothing(SG)and Standard Normal Variate Transformation(SNV),Partial least square-discriminant Analysis(PLS-DA),Random Forests(RF),and Support Vector Machine(SVM)and other methods are used to establish the classification model.The accuracy of prediction results is used as the evaluation index to select the best model.A confusion matrix was used to evaluate model classification performance.The results showed that the Multiplicative scatter correction(MSC)combined with the PLS-DA method was the optimal identification model for outer epidermis data,and the identification accuracy of the prediction set reached 97.59%.For inner epidermis data,the raw data combined with the PLS-DA method was used as the optimal identification model,and the identification accuracy of the predictio

关 键 词：高光谱技术 模式识别 新会陈皮:陈化年份 

分 类 号：TS272.7[农业科学—茶叶生产加工]