采用衰减全反射傅里叶变换红外光谱(FTIR-ATR)结合模式识别的方法建立橄榄油的快速鉴别模型。实验中测定不同品牌的纯橄榄油样本与纯大豆油样本以及橄榄油中掺入大豆油的混合油样本的FTIR-ATR 光谱。将625~4000 cm-1波段的光谱数据作为鉴别研究对象，应用主成分分析法(PCA)对光谱数据进行降维处理，再结合Fisher 判别分析和多层感知器神经网络两种模式识别的方法对纯橄榄油、纯大豆油以及三种掺杂有大豆油的橄榄油样本进行鉴别。结果表明：对原始数据所采用的预处理方法中，经小波去噪后的数据的鉴别效果最佳。PCA 结合Fisher 判别分析建立的判别模型，原始判别的准确率达到100%，交叉验证的准确率达到97.1%；多层感知器神经网络鉴别模型训练集分类准确率为100%，测试集分类准确率为100%。因此，FTIR-ATR 光谱技术结合模式识别方法可以实现对纯橄榄油、纯大豆油以及三种掺杂有大豆油的橄榄油样本的鉴别，而且具有操作简便、快速、可靠等优点。

Attenuated total reflectance-Fourier transform infrared spectroscopy (FTIR-ATR) combined with pattern recognition is applied to establish a rapid discrimination model of olive oil. Measurements of FTIR-ATR spectra of pure olive oil samples and pure soybean oil samples from different brands and mixed samples of pure olive oil doped with soybean oil are performed. The obtained spectra between 625 cm-1 and 4000 cm-1 are regarded as the research object of discrimination analysis. The principal component analysis (PCA) is applied to the dimension reduction of these spectra. Identification of the pure olive oil samples, pure soybean oil samples and three samples of olive oil doped with soybean oil is carried out by two pattern recognition methods, the Fisher discrimination and multilayer perception neural network. The experiment results show that the identification effect after the wavelet denoising is the best among all the adopted pretreatment methods. The original discriminant accuracy of discrimination model established by PCA combined with Fisher discriminant analysis reaches 100%. The cross validation accuracy reaches 97.1%. Classification accuracy of the training set and the test set by multilayer perception neural network identification model is 100% . Therefore, FTIR-ATR spectroscopy combined with pattern recognition method can realize the identification of the pure olive oil samples, pure soybean oil samples and three different oil blend samples. This method has some advantages over other methods owing to its easy operation, high speed and reliability.