Classification of Apple Varieties Using FT-NIR Spectroscopy and Possibilistic Learning Vector Quantization

Abstract

Red Fuji, Huaniu and Gala were classified by Fourier transform near infrared (FT-NIR) spectroscopy and possibilistic learning vector quantization (PLVQ) which was proposed to solve the noise sensitivity problem of fuzzy learning vector quantization (PLVQ). Firstly, apple NIR spectra were measured by FT-NIR spectrophotometer. Secondly, principal component analysis (PCA) was used to compress the dimensionality of NIR spectra which was high dimensional. Thirdly, fuzzy c-means (FCM) clustering was run to termination to obtain the cluster vectors for PLVQ. Finally, PLVQ was performed to classify the data. Experimental results showed that this classification method was fast, nondestructive and effective for classifying the variety of apples.