Abstract: Classification of apples is an important link in postharvest commercialization processing. To realize the non-
destructive, rapid and effective discrimination of apple fruits, the near infrared reflectance spectra of four varieties of apples are collected using near infrared spectroscopy, reduced by Principal Component Analysis（PCA） and used to extract the discriminant information by Linear Discriminant Analysis（LDA）, Quadratic Discriminant Analysis（QDA）, Fuzzy Uncorrelated Discriminant Transformation（FUDT） and Foley-Sammon discriminant analysis. K-nearest neighbor is used to finish the classification. The classification results show that FUDT can extract the discriminant information of NIR spectra more effectively, and achieves the highest classification accuracy.

Key words: classification of apples, near infrared spectroscopy, Linear Discriminant Analysis（LDA）, Quadratic Discriminant Analysis（QDA）, Fuzzy Uncorrelated Discriminant Transformation（FUDT）, Foley-Sammon discriminant analysis

摘要： 苹果的分类是苹果采收后商品化处理的重要环节。为了快速、无损和有效地实现苹果的分类，利用近红外光谱技术采集四种苹果的近红外反射光谱，用主成分分析对高维的近红外光谱进行降维处理，分别运行线性判别分析，二次判别分析，模糊非相关判别转换和Foley-Sammon判别分析提取鉴别信息，用k-近邻分类器进行分类。分类结果表明，模糊非相关判别转换能更好地提取苹果近红外光谱的品种鉴别信息，达到了最高的分类准确率。

关键词: 苹果分类, 近红外光谱, 线性判别分析, 二次判别分析, 模糊非相关判别转换, Foley-Sammon判别分析