以四种品牌152组食用醋样品为研究对象, 采用漫反射与透射两种近红外光谱采集模式分别进行光谱数据采集, 并以此建立了食用醋品牌溯源模型, 重点考察光谱采集模式、 光谱预处理方法等对溯源模型精度的影响。 结果表明, 选取114组样品为训练集, 原始光谱数据经过多元散射校正、 二阶求导预处理后, 采用偏最小二乘判别分析法(PLS1-DA)建立的食用醋NIRS品牌溯源模型, 对38组测试集样品进行预测, 透射光谱模型的决定系数(R2)、 校准均方根误差(root-mean-square error of calibration, RMSEC)、 预测均方根误差(root-mean-square error of prediction, RMSEP)分别为0.92, 0.113, 0.127, 正确识别率为76.32%; 漫反射光谱模型R2, RMSEC, RMSEP分别为0.97, 0.102, 0.119, 正确识别率为86.84%。 由此说明, 近红外光谱结合PLS1-DA可以用来建立食用醋品牌溯源模型, 且漫反射光谱模型预测效果更好。

In the present paper, 152 vinegar samples with four different brands were chosen as research targets, and their near infrared spectra were collected by diffusion reflection mode and transmission mode, respectively. Furthermore, the brand traceability models for edible vinegar were constructed. The effects of the collection mode and pretreatment methods of spectrum on the precision of traceability models were investigated intensively. The models constructed by PLS1-DA modeling method using spectrum data of 114 training samples were applied to predict 38 test samples, and R2, RMSEC and RMSEP of the model based on transmission mode data were 0.92, 0.113 and 0.127, respectively, with recognition rate of 76.32%, and those based on diffusion reflection mode data were 0.97, 0.102 and 0.119, with recognition rate of 86.84%. The results demonstrated that the near infrared spectrum combined with PLS1-DA can be used to establish the brand traceability models for edible vinegar, and diffuse reflection mode is more beneficial for predictive ability of the model.