基于近红外光谱和OPLS-DA的不同牌号卷烟分类识别方法研究

为了对卷烟牌号进行准确分类鉴别，提出了一种基于近红外光谱(NIRS)分析技术结合有监督的模式识别快速鉴别卷烟牌号的新方法。利用标准正态变量变换(SNV)、多元散射校正(MSC)、一阶导数(FD)、二阶导数(SD)和Savitzky-Golay平滑(SG)及其相结合的光谱预处理方法对烟丝光谱进行预处理，通过近红外光谱结合主成分分析(PCA)、偏最小二乘判别分析(PLS-DA)和正交偏最小二乘判别分析(OPLS-DA) 3种模式识别方法对不同牌号烟丝进行分类识别研究，并采用分类识别正确率作为评价指标。实验结果表明：(1)烟丝近红外光谱主成分得分图交叉重叠，区分不明显，PCA无法识别出5种牌号的成品烟丝；(2)烟丝光谱经MSC+FD预处理后的PLS-DA模型可得到较好的识别效果，校正集和测试集的分类识别正确率分别为100%和98.3%；(3)烟丝光谱经MSC+SD预处理后的OPLS-DA模型的模式识别效果最好，模型对自变量拟合指数(R2X），因变量的拟合指数(R2Y)和模型预测指数(Q2)分别为0.485、0.907 和0.748，近红外光谱校正集和测试集的分类识别正确率均为100%。说明近红外光谱技术结合有监督模式识别方法OPLS-DA建立的烟丝牌号分类模型具有高效快速、准确无损的优点，为卷烟烟丝分类提供了一种新的快速鉴别方法。

Classification Discrimination of Different Types of Cigarette Based on Near Infrared Spectroscopy and OPLS-DA Algorithm

A novel near infrared spectroscopy(NIRS) combined with supervised pattern recognition was proposed for the rapid classification and discrimination of types of cigarettes.Standard normal variables(SNV),multiplicative signal correction(MSC),first derivative(FD),second derivative(SD) and Savitzky-Golay filt(SG) and their combined spectral pre processing methods were used for the spectral data preprocessing of finished cut tobacco.A discriminant model was established by NIRS combined with three pattern recognition methods include principal component analysis(PCA),partial least squares discriminant analysis(PLS-DA) and orthogonal partial least squares discriminant analysis(OPLS-DA),and the prediction accuracy of classification identification was used as an evaluation index.The experimental results showed that:(1) the principal component distribution maps were intertwined,and PCA could not identify five types of finished cut tobacco.(2) The PLS-DA model for finished cut tobacco spectrum after MSC+FD pretreatment could achieve better classification and recognition results,and the prediction accuracies for the calibration set and test set were 100% and 98.3%,respectively.(3) The identification of the OPLS-DA model for the finished cut tobacco spectrum after MSC+SD pretreatment was the best,and the parameters of the model,including the fraction of the variation of X explained by the model(R2X),the fraction of the variation of Y explained by the model(R2Y),and the fraction of the variation of Y that can be predicted by the model according to the cross validation(Q2) were 0.485,0.907 and 0.748,respectively.The prediction accuracies for the calibration set and test set both reached to 100%.Results showed that the classification model based on NIRS combined with OPLS-DA was efficient,quick,accurate and non destructive,and provided a new and rapid identification approach for finished cut tobacco classification.