Abstract: The model of discrimination analysis was established by PLS-Mahalanobis distance for differentiating raw milk and adulterated milk which was added with vegetable cream. Three principal components were compressed from original data by PLS. The whole wave numbers regions were selected without any pretreatment methods,and all the accuracy rates of this model are 100%. Meanwhile,10 unknown samples were used to predict the results by the model,and all the prediction accuracy rates are 100%. PLS model for detecting the content of vegetable cream added with raw milk was set up with good veracity. Reliability of the model was verified by cross-validation and external-validation. Predictive correlation coefficient of the content of vegetable cream by PLS model is 0.9963. Root mean square error of calibration(RMSEC) is 0.11,while root mean square error of cross validation(RMSECV) is 0.142. By paired samples test,the results of prediction are compared with real values with no significantd ifference.

Key words: near infrared spectroscopy, pattern recognition, partial least squares, raw milk, adulterated milk