基于提取不同中红外光谱特征信息的烟叶部位判别研究
- 分享到:
摘要:
中红外光谱(MIR)分析技术在烟草中有广泛的应用,利用中红外分析可以获取烟草中大量化学信息.为了提高谱图的信噪比,需要对谱图数据进行预处理.研究发现对烟叶中红外光谱数据进行一阶导数结合Savizky-Golay的预处理,不仅提高了信噪比,而且增加了烟叶部位分类判别的准确率.另外,对谱图数据进行降维处理,有利于提取中红外谱图信息,减少冗余数据,减少计算时间.本文对比了基于原始中红外谱图数据、连续投影算法(SPA)特征提取后数据、偏最小二乘法(PLS)降维特征提取后数据的烟叶部位分类判别准确率,结果表明PLS降维特征提取可以有效提取烟叶中红外光谱数据的特征信息,有利于烟叶部位分类判别准确率的提高.利用PLS提取烟叶中红外特征信息数据建立的烟叶部位支持向量机(SVM)分类判别模型,其建模、留一法和独立测试集的准确率分别为:96.00%、89.60%和80.65%.
Mid-infrared spectroscopy(MIR)technique has been widely applied in tobacco analysis,which could be used to obtain numerous chemical information.To investigate the IR spectrum data and improve the S/N value that contains large amount of tobacco information,data pre-treatment must be conducted.In this work,first-order derivative coupled with Savizky-Golay data treatment can not only improve the S/N value but also result in a better prediction accuracy of support vector classification(SVC)for tobacco leave part.It is found that dimension reduction can be used to eliminate redundancy information which helps to extract the characteristic data and saves the computation time.Here,Partial least squares(PLS)and successive projection algorithm(SPA)were used to extract the characteristic data from the infrared spectrum.The prediction accuracy of SVC for tobacco leave part showed that the result of dimension reduction via the PLS algorithm was better than that via the SPA algorithm.It can be concluded that the prediction accuracy of SVC for tobacco leave part was improved by using dimension reduction via the PLS to extract the most valuable data from MIR.In a word,the SVC accuracy for tobacco leave part from PLS data of training set,leave one out cross validation and testing set were 96.00%,89.60%and 80.65%,respectively.
作者:
赵娟娟 叶顺 徐可 陈栋骅 岳宝华 李敏杰 刘太昂 陆文聪
Zhao Juanjuan;Ye Shun;Xu Ke;Chen Donghua;Yue Baohua;Li Minjie;Liu Taiang;Lu Wencong(Department of Chemistry,Shanghai University,Shanghai 200444,China;Department of Electronic Information Materials,Shanghai University,Shanghai 200444,China;Technology Center of Shanghai Tobacco Group Co.,Ltd.,Shanghai 200082,China)
机构地区:
上海大学化学系 上海大学电子信息材料系 上海烟草集团有限责任公司技术中心
出处:
《betway官方app 学报：自然科学版》 CAS 北大核心 2021年第1期45-49,共5页
基金:
国家自然科学基金青年基金(21706156) 卷烟烟气重点实验室开放性课题(K2018-1-056P).
关键词:
中红外光谱 连续投影算法 支持向量机 烟叶部位
MIR SPA SVM tobacco leave parts
分类号:
O69 [理学—化学]
