利用可见―近红外光谱术无损检测牛奶中的三聚氰胺

Nondestructive Detection of Melamine of Milk Using VIS-NIR Spectroscopy

初步探讨了利用可见一近红外光谱术检测牛奶中三聚氰胺的可行性及方法。通过往牛奶中掺人不同浓度三聚氰胺的方法．制备了165个样品，三聚氰胺浓度为0-1000ppm。利用光纤光谱仪采集样本的可见一近红外光谱。其光谱范围为350。1800nm。然后分别采用最小二乘法（PtS）、区间偏最小二乘法（IPLS）及联合区间偏最小二乘法（siPtS）建立预测模型。比较实验结果表明：把光谱分为10个子区间，通过SIPLS方法。选出3个光谱子区间（4、7、9）联合建立的预测模型最优．其校正集和预测集得相关系数分别为0．9981和0．9946．校正集和预测集的均方根误差分别为0．1942和0．3299。因此，可见近红外光谱术结合联合区间偏最小二乘法能无损、快速的检测牛奶中的三聚氰胺。

The study discussed how visible-near infrared spectroscopy （Vis-NIRS） can be used to detect the melamine in the milk. Through adulterating 0-1000ppm content of melamine into pure milk, 165 samples were prepared in all. The spectra are recorded in the Vis-NIR range from 350 to 1800nm using the fiber spectrometer. Partial least square （PLS）, interval partial least square （IPI.S） and synergy interval partial least square （SIPLS） methods were used to establish the calibration models of melamine. The results showed that the optimal SIPLS model is obtained with 10 intervals and the optimal combinations of three intervals are number g, 7, 9. The Correlation coefficient （r） between the predicted and actual melamine is 0.9981 and 0.9946 for calibration and prediction set, respectively. The root mean square error of cali- bration （RMSEC） and prediction （RMSEP） set was 0.1942 and 0.3299, respectively. In conclusion, the combination of Vis-NIR spectroscopy and SIPLS methods can be used to provide a technique of noninvasive and rapid analysis for melamine in milk.