Classification of Apple Varieties Using Near Infrared Reflectance Spectroscopy and Fuzzy Discriminant C‐Means Clustering Model

Abstract

Abstract This research proposed a novel fuzzy clustering model, called fuzzy discriminant c‐means (FDCM) clustering, to classify apple varieties couple with near infrared spectroscopy. FDCM can not only partition samples but also extract discriminant information from them by introducing fuzzy linear discriminant analysis (FLDA) into fuzzy c‐means (FCM) clustering. The 200 apple samples of four apple varieties were made experiments to collect the near infrared reflectance (NIR) spectra. Principal component analysis (PCA) was used to reduce the high dimensionality of NIR spectra data. FDCM and FCM were performed as classifiers to cluster the data, respectively. The clustering accuracy of FDCM achieved 97% which was higher than FCM, possibilistic c‐means and Gustafson–Kessel (GK) clustering. Furthermore, soluble solids content and total acidity of apples were measured, and the quantitative results showed that different types of apples contain different content of the nutritional composition. This makes it feasible for classification of apple varieties via NIR technology. The clustering results obtained in experiments showed that NIR spectroscopy combined with PCA and FDCM clustering could successfully discriminate apple varieties. Practical Applications Postharvest processing of apple is very important for the agricultural industry. For this purpose, the proposed fuzzy discriminant c‐means clustering model coupled with near infrared spectroscopy can be used to select higher quality of apples for grading and sorting machines.