中文摘要：

运用近红外光谱技术结合模式识别方法对液态奶中违法添加三聚氰胺进行快速检测。采集纯牛奶以及9种不同质量浓度掺假三聚氰胺牛奶的近红外光谱图，运用主成分分析方法、线性判别分析方法以及基于虚拟矢量编码的偏最小二乘判别分析方法对纯牛奶和不同质量浓度掺假三聚氰胺的牛奶近红外光谱数据进行判别分析。结果表明，相比主成分分析模型和线性判别分析模型，基于虚拟矢量编码的偏最小二乘判别分析模型的训练和预测性能均最好，识别训练集和预测集正确率能分别达到100％和90．32％。该法简单、快速、准确，为客观评价食品质量等提供了一种新的可选的方法。

英文摘要：

In this paper, near-infrared （NIR） spectroscopy fiber technology combined with pattern recognition was used for rapid identifica- tion of me/amine adulteration in milk.The fingerprint information of the pure milk and different melamine-adulterated milk were obtained by NIR fiber technology, subsequently, pattern recognition models including principal component analysis （PCA）, hnear discriminant analysis （LDA） and partial least squares discriminant analysis（PLSDA） based on dummy code of vector were employed to identify the pure milk and different melamine-adulterated level in milk. The obtained results indicated that PLSDA model can achieve a best satisfactory recognition performance compared with PCA and LDA models. Recognition rates of PLSDA model for train set and prediction are 100% and 90.32%, re- spectively.Therefore, the proposed method holds great potential to be extended as a promising alternative for more apphcations in food quali- ty control field.