| 【Objective】 As precursor substances of flue gas components, volatile aroma components can affect
the quality of tobacco sensory evaluation. The aims of this study were to analyze the content of volatile compounds in fluecured tobacco samples from Ganzhou, to explore the content regulation of volatile compounds in tobacco leaves of different producing areas, and to predict the origin classification of samples so as to achieve quality control of tobacco. 【Methods】
54 kinds of volatile compounds of 62 flue-cured tobacco samples from 6 tobacco-producing counties(Shicheng, Ruijin,
Anyuan, Huichang, Xingguo, Xinfeng) in Ganzhou were measured by static headspace-gas chromatography-mass spectrometry
method(HS-GC-MS). Subsequently, principal component analysis(PCA) was used for classification of sample origins, genetic
algorithm (GA) was used for variable screening, and three supervised pattern recognition methods, namely, linear discriminant
analysis (LDA), back propagation-artificial neural network (BP-ANN) and least squares-support vector machines (LS-SVM)
were used for origin prediction. 【Results】 The classification effect by PCA, which was contributed to explore the content
regulation of volatile compounds of samples in Ganzhou, was significantly improved after obtaining 11 volatile compounds
(variables) by GA. The prediction accuracy rates of BP-ANN and LS-SVM reached 96.8% and 98.4%, respectively, which
showed good practical application value. On this basis, the identification model for the origin of flue-cured tobacco of Ganzhou
was established. 【Conclusion】 The method proposed in this paper can accurately determine the origins of flue-cured tobacco
samples from Ganzhou, find the content regulation of volatile compounds among different origins, and be used for the monitoring
of flavor patterns and the quality control of flue-cured tobacco across the country. |
