近红外光谱法用于掺假羊奶的快速无损鉴别

Rapid and nondestructive identification of adulteration Goat's milk by near infrared Spectroscopy

利用近红外光谱技术结合多种化学计量学方法,研究了快速鉴别掺假羊奶的方法。将淀粉溶液,含尿素的淀粉溶液,含尿素和奶油的淀粉溶液按不同比例掺入纯羊奶中,进行近红外光谱采集。分别采用偏最小二乘差别分析(PLS-DA),fisher线性判别和多层感知器(MLP)神经网络法建立校正模型并进行检验验证。结果表明,MLP神经网络的鉴别效果最好,其校正模型的正判率达到99.4%,验证集的正判率达到100%。说明采用近红外光谱技术结合适当的化学计量学方法可以实现羊奶掺假检测的快速无损鉴别。

Rapid and nondestructive identification of adulteration goat＇s milk was studied by the the combination of near infrared spectroscopy and chemometric methods. Adulteration samples of starch solution, starch solution contained urea, starch solution contained urea and margarine were collected and scanned by near infrared spectrograph. Using Partial least squares difference analysis （PLS-DA）, fisher＇s linear discriminator and Muldlayer Perceptron （MLP） artificial neural network to built calibration models, then the models were verified. The results showed that the highest rate of correct discriminant of the models built by MLP artificial neural network was 99.4%, and the prediction accuracy was 100%.All of these showed that the near infiared spectnam technique combined with chemometric methods could be used for the rapid and nondestructive identification of adulteration goat＇s milk.