中文摘要：

提出了一种衰减全反射红外光谱法快速分类和识别多种食用油的方法——KL-BP模型。此模型利用KL算法对原始光谱数据分类特征进行提取并对原始数据降维,降维后的数据作为神经网络的输入建立分析模型。实验共收集了九种食用油包括芝麻油、玉米油、油菜籽油、调和油、葵花油、花生油、橄榄油、大豆油、茶籽油,共84个样品,并测定了其衰减全反射红外光谱。为了对比所提方法性能,分别建立PCA直接分类、KL直接分类、PLS-DA、PCA-BP和KL-BP模型的分类结果进行对比。研究结果表明,对所研究的9种食用油,PCA直接分类、KL直接分类、PLS-DA、PCA-BP和KL-BP方法的识别率分别为59.1%,68.2%,77.3%,77.3%和90.9%。在数据降维中,KL算法通过分别提取使类间距离和类内距离比值最大方向的特征向量提取和包含在类内离散度矩阵中的分类信息,能够比PCA方法提取了更多的分类信息;引入BP神经网络能有效地提高分类能力和分类准确率;KL-BP综合了KL对分类信息提取优势以及BP神经网络自学习、自适应、非线性的优点,在分类和识别成分相近的9种食用油中表现出了最优秀的能力。

英文摘要：

A rapid discrimination method of edible oils,KL-BP model,was proposed by attenuated total reflectance infrared spectroscopy.The model extracts the characteristic of classification from source data by KL and reduces data dimension at the same time.Then the neural network model is constructed by the new data which as the input of the model.84 edible oil samples which include sesame oil,corn oil,canola oil,blend oil,sunflower oil,peanut oil,olive oil,soybean oil and tea seed oil,were collected and their infrared spectra determined using an ATR FT-IR spectrometer.In order to compare the method performance,principal component analysis（PCA）direct-classification model,KL direct-classification model,PLS-DA model,PCA-BP model and KL-BP model are constructed in this paper.The results show that the recognition rates of PCA,PCA-BP,KL,PLS-DA and KL-BP are 59.1%,68.2%,77.3%,77.3% and 90.9%for discriminating the 9kinds of edible oils,respectively.KL extracts the eigenvector which make the distance between different class and distance of every class ratio is the largest.So the method can get much more classify information than PCA.BP neural network can effectively enhance the classification ability and accuracy.Taking full of the advantages of KL in extracting more category information in dimension reducing and the features of BP neural network in self-learning,adaptive,nonlinear,the KL-BP method has the best classification ability and recognition accuracy and great importance for rapidly recognizing edible oil in practice.