Classification Technique for Using Near Infrared Spectroscopy to Detect Dry Juice Sac of Mandarin Fruit cv. Sai Nam Pueng

Abstract

The use of chemometrics to analyze NIRs spectral data of Mandarin fruit Citrus reticulata cv. Sai Nam Pueng for dry juice sac detection was investigated. The total of 200 fruits (75 normal and 125 abnormal fruits) were acquired the spectral data using NIRSystem 6500 with fiber optic probe in the wavelength region from 700–1100 nm prior to determine the reference laboratory data including moisture content (MC), total soluble solid (TSS) and titratable acidity (TA). Then, the principal component analysis (PCA) and multiple linear regression - discriminant analysis (MLR-DA) were used to classify fruit with this symptom. It was found that Savitzky-Golay smoothing (10 nm average for left and right sides) and second derivative (10 nm average for left and right sides) were used as spectra pre-treatment method for the optimal PCA and MLR-DA model performances of MC, TSS and TA. The effective wavelengths in MLR-DA were at 826, 924, 962, 978, 1008 and 1028 nm. The precision of MLR-DA was 96%. There were only 4% of un-classify samples which were 1–25% of dry juice sac fruits and normal fruits. Therefore, MLR-DA could be the efficient method to classify a dry juice sac of Mandarin using non-destructive technique NIRs.