Abstract: NIR spectroscopy analysis method is applied to classify different kinds of apple samples. A new method of apple NIR spectra qualitative analysis based on uncorrelated discriminant transformation is presented. Three methods of feature extraction such as principal component analysis, Fisher discriminant analysis and uncorrelated discriminant transformation are used to extract feature from apple NIR spectra. By using K-nearest neighbour（KNN） three classification models are constructed for the classification recognition of apples. The models are verified by using the leave-one-out cross-validation. The results show that Uncorrelated Discriminant Transformation（UDT） model, comparing with Principal Component Analysis（PCA） model and Fisher Discriminant Analysis（FDA） model, has higher accuracy recognition rate.

Key words: Near Infrared（NIR） spectroscopy, principal component analysis, Fisher discriminant analysis, uncorrelated discriminant transformation, feature extraction

摘要： 采用近红外光谱分析法对不同种类的苹果样品进行分类，提出一种基于非相关判别转换的苹果近红外光谱定性分析新方法。实验分别采用主成分分析、Fisher判别分析和非相关判别转换三种方法对苹果光谱数据进行特征提取，并使用K-近邻分类算法建立三种苹果分类识别模型，最后使用“留一”交叉验证法进行模型检验。结果表明，使用非相关判别转换方法建立的模型正确识别率优于使用主成分分析和Fisher判别分析建立的模型。

关键词: 近红外光谱, 主成分分析, Fisher判别分析, 非相关判别转换, 特征提取