Non-destructive identification and estimation of granulation in 'sai Num Pung’ tangerine fruit using near infrared spectroscopy and chemometrics

• Main Authors: Parichat Theanjumpol, Kumpon Wongzeewasakun, Nadthawat Muenmanee, Sakunna Wongsaipun, Chanida Krongchai, Viboon Changrue, Danai Boonyakiat, Sila Kittiwachana
• Format: Journal
• Published: 2019
• Subjects: Agricultural and Biological Sciences
• Online Access: https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85063317255&origin=inward; http://cmuir.cmu.ac.th/jspui/handle/6653943832/65241
• Institution: Chiang Mai University

© 2019 Elsevier B.V. Granulation or ‘dry juice sac’ is a physiological disorder, which has a negative effect on the eating quality of citrus fruit. It is not easy to identify the fruit with dry juice sacs until the peel is removed. This research describes a quick and non-destructive method for detecting and estimating the occurrence of granulation in 'sai Num Pung’ tangerine based on the use of near infrared (NIR) spectroscopy and chemometric analysis. NIR spectra of 178 fruit samples were recorded after harvest and one day of storage at 25 °C. Moisture content (MC), soluble solids content (SSC) and titratable acidity (TA) were analyzed. The fruit were rated into five classes from A (no visible of granulation) up to E (most of the fruit body was opaque or the estimated percentage of granulation was more than 75%). Partial least squares (PLS) regression was used to investigate the relationship between the quality parameters and the occurrence of the granulation disorder. Classification models such as linear discriminant analysis, quadratic discriminant analysis, partial least squares-discriminant analysis, k nearest neighbor and supervised self-organizing map (SSOM) were used to identify the granulation classes. The predictive results from PLS modelling revealed that the disorder could be related to lower MC, SSC and TA of the fruit. The results of this analysis supported the idea that spectroscopic measurement could be used to assess the incidence of granulation externally. In this research, SSOM, as a representative of non-linear classification, resulted in the best classification performance where the percentages of predictive ability, model stability and correctly classified (CC) were 93.7%, 95.3% and 94.0%, respectively, for the test samples generated by bootstrap method. The SSOM model was also tested with external validation samples which were the tangerine harvested in different season resulting in the CC of 78.4%.