结合SG-LDA-SVC的太赫兹时域光谱快速无损鉴别陈皮产地的方法研究  

Rapid Nondestructive Identification of Tangerine Peel Origin based on Terahertz Time-domain Spectroscopy combined with SG-LDA-SVC

作　　者：余静孝 韦庆益[1] 罗洁 李茜 蒲洪彬[1] Yu Jingxiao;Wei Qingyi;Luo Jie;Li Qian;Pu Hongbin(School of Food Science and Engineering,South China University of Technology,Guangzhou 510641,China;China Communication Technology(Jiang Men)Corporation,Jiangmen 529000,China)

机构地区：[1]华南理工大学食品科学与工程学院,广东广州510641 [2]江门市华讯方舟科技有限公司,广东江门529000

出　　处：《现代农业装备》2022年第4期16-21,28,共7页Modern Agricultural Equipment

基　　金：广东省乡村振兴战略专项项目(2022KJ101)。

摘　　要：由于陈皮品质与产地关系密切,陈皮的道地性检测显得尤为重要。本文利用太赫兹时域光谱技术对不同年份与产地的陈皮进行快速无损检测方法研究。首先,获取2017、2018、2019、2020和2021年5个年份陈皮(包括古井、七堡、双水,梅江和三江5个产地)的太赫兹光谱,经多种预处理方法处理,建立了K邻近分类(KNNC)、支持向量机分类(SVC)和随机森林分类(RFC)的陈皮鉴别模型。陈皮太赫兹光谱经卷积平滑(SG)预处理后,结合线性判别分析(LDA)方法提取太赫兹光谱特征的SVC模型(SG-LDA-SVC),测试集与预测集的准确度均达100%。太赫兹时域光谱结合化学计量学为陈皮产地快速无损检测提供了一个有潜力的新方法。Due to the close relationship between the quality of tangerine peel and its producing area,the authentic detection of tangerine peel is particularly important.In this paper,terahertz time-domain spectroscopy was used to detect tangerine peel of different years and places of origin.Firstly,the terahertz spectra of tangerine peel(gujing,qipu,shuangshui,meijiang and sanjiang)of five years in 2017、2018、2019、2020 and 2021 were obtained and processed by various pretreatment methods.The identification models of Tangerine peel based on K neighborhood classification(KNNC),support vector machine classification(SVC)and random forest classification(RFC)were established.The SVC model(SG-LDA-SVC)was used to extract terahertz spectral features of Orange cortex after convolution smoothing(SG)pretreatment.The accuracy of test set and prediction set was 100%.In this paper,SG-LDA-SVC model based on THZ-TDS provides a potential new method for rapid nondestructive testing of tangerine peel production area.

关 键 词：陈皮产地 太赫兹 化学计量学 快速检测 

分 类 号：TS273[轻工技术与工程—农产品加工及贮藏工程]