基于近红外光谱的大鲵肉粉掺伪鉴别及纯度检测

利用近红外光谱技术进行大鲵肉粉的掺伪鉴别及纯度检测。分别采集大鲵纯肉粉、掺入江团鱼肉粉、草鱼肉粉和土豆淀粉的掺伪大鲵肉粉(各40个样本,4类共160个样本)的近红外光谱图。原始光谱经光谱预处理后,利用偏最小二乘-判别分析(partial least square-discriminant analysis,PLS-DA)法分别建立2分类(纯样和掺伪样)和4分类(纯样、掺江团鱼样、掺草鱼样和掺淀粉样)的定性判别模型,利用偏最小二乘回归(partial least squares regression,PLSR)分析法分别建立3类掺伪大鲵肉粉的纯度定量校正模型。结果表明,PLS-DA定性模型中,经一阶导数+多元散射校正光谱预处理后,所建2分类和4分类模型性能均为最佳,校正集和预测集的预测准确率均为100%;PLSR定量模型中,大鲵肉粉掺江团鱼肉粉、大鲵肉粉掺草鱼肉粉和大鲵肉粉掺土豆淀粉模型的校正集相关系数(R_c~2)分别为0.990 6、0.986 4和0.993 3,校正集的均方根误差分别为1.14%、1.39%和0.88%;测试集的相关系数(R_p~2)分别为0.994 4、0.992 4和0.990 8,测试集的均方根误差分别为0.83%、0.89%和1.22%。运用近红外光谱技术结合化学计量学方法能够对大鲵肉粉进行掺伪鉴别及纯度检测。

Adulteration and Purity Detection of Chinese Giant Salamander Meat Powder Based on Near Infrared Spectroscopy

In this paper, the adulteration and purity detection of Chinese giant salamander (Andrias davidianus) meat powder were investigated based on near infrared spectroscopy. Infrared spectra of authentic Chinese giant salamander meat powder and adulterated samples with longsnout catfish meat powder, grass carp meat powder or potato starch (40 samples each, 160 samples in total) were collected. After spectral preprocessing, qualitative discrimination models between pure and adulterated samples and among pure sample and adulterated samples with longsnout catfish, grass carp and potato starch) were established by partial least square discriminant analysis (PLS-DA). Partial least squares regression (PLSR) was used to establish a quantitative calibration model for the purity of the three adulterated samples. The results showed that the PLSDA qualitative models with first derivative + multiplicative scatter correction (MSC) spectral preprocessing had the best performance in discriminating the authentic and adulterated samples with 100% predictive accuracy for the calibration and prediction sets. For the PLSR quantitative models of adulteration of longsnout catfish, grass carp and potato starch, the correlation coefficients of the calibration set (Rc 2) were 0.990 6, 0.986 4, and 0.993 3, respectively, and the root mean square errors of calibration (RMSEC) were 1.14%, 1.39%, and 0.88%, respectively. The correlation coefficients of the prediction set (Rp 2) were 0.994 4, 0.992 4, and 0.990 8, respectively, and the root mean squared error of prediction (RMSEP) was 0.83%, 0.89%, and 1.22%, respectively. These results suggest that near infrared spectroscopy combined with chemometrics could be used for adulteration and purity detection of Chinese giant salamander meat powder.