CAJ | 학술논문

目的：研究烤烟香气风格中焦香、辛香、甜香等香韵的识别技术。方法：采用近红外光谱技术结合最小二乘支持向量机（L S-S V M）模式识别方法。烟叶粉末的近红外漫反射光谱经过波长范围选择和多种预处理优化后输入模型，使用k折交互验证和多层网格法优化LS-SVM模型参数，建立三种香韵识别模型。结果：焦香、甜香、辛香的识别准确率CR分别为94.7%、88.9%、94.8%，ROC曲线下面积AUC分别为0.99、0.99、1.00。结论：说明使用近红外光谱技术结合LS-SVM方法可有效识别烤烟香气风格。
목적：연구고연향기풍격중초향、신향、첨향등향운적식별기술。방법：채용근홍외광보기술결합최소이승지지향량궤（L S-S V M）모식식별방법。연협분말적근홍외만반사광보경과파장범위선택화다충예처리우화후수입모형，사용k절교호험증화다층망격법우화LS-SVM모형삼수，건립삼충향운식별모형。결과：초향、첨향、신향적식별준학솔CR분별위94.7%、88.9%、94.8%，ROC곡선하면적AUC분별위0.99、0.99、1.00。결론：설명사용근홍외광보기술결합LS-SVM방법가유효식별고연향기풍격。
A method combined with near infrared (NIR) spectroscopy and least squares-support vector machine (LS-SVM) was applied to study identiifcation technology of tobacco aroma styles. The NIR spectrum of the tobacco powder were preprocessed by a wavelength selection technique and several pretreatment methods including smoothing, multiplicative scatter correction and standard normal variate transformation. The LS-SVM identiifcation models for three kinds of tobacco aroma styles were built, after optimizing parameters by k-fold cross validation and multilayer grid search. The values of accuracy rate of burnt aroma, spice aroma and sweetness aroma model were 94.7%, 88.9% and 94.8%, respectively. And the area under AOC curve were 0.99, 0.99 and 1.00, respectively. The overall results show that NIR spectroscopy combined with LS-SVM can be efifciently utilized for rapid and accurate identiifcation of tobacco aroma styles.