The evaluation and classification of carambola using portable near infrared spectrometer (NIR)

Author(s)

Ingrid Alves de Moraes

• Document type: Master's Dissertation
• Press: Campinas, SP.
• Institution: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia de Alimentos
• Defense date: 2021-08-10

Examining board members

Douglas Fernandes Barbin; Daniela Souza Ferreira; Raul Benito Siche Jara

• Advisor: Douglas Fernandes Barbin

Abstract

Averrhoa carambola is known as star fruit, due to its shape. It is consumed in high quantities in Asia and South and Central America as it is a source of vitamins and minerals. The classification of carambola is carried out by trained evaluators through visual inspections. It is an expensive and subjective method, which can generate high variability in the results, since the repetitive analysis of thousands of samples, makes it difficult to standardize this task. Thus, this project aims to use near-infrared spectroscopy (NIR) as a non-destructive method for predicting and classifying carambola maturity stages. First, the absorbance spectra were obtained on a portable NIR spectrometer. Then, the colour and composition characteristics of soluble solids, vitamin C, moisture, pH and titratable acidity were determined by reference analytical methods. Multivariate analyses such as principal component analysis (PCA), partial least squares regression (PLSR) partial least squares and discriminant analysis (PLS-DA) were used to predict the physical and physical-chemical parameters of the samples. Pre-treatment was tested in spectra, including first and second derivatives, to reduce variations caused by factors unrelated to composition and improve the accuracy of statistical models. The selection of variables by Interative partial least square (iPLS) and Genetic algorithm (GA-PLS) improved the performance of the PLSR models, reaching (R2P of 0.78 and 0.74), (RMSEP of 0.2 and 0.87), (RPD of 2, 01 and 2.23) and (RER of 8.02 and 10.38), for pH and moisture content, respectively, which suggests to be acceptable models for quality control with quantitative predictions with low errors. The PLS-DA model with variable selection by iPLS obtained the best performance with 84.2 % accuracy to classify a carambola according to the stage of maturity. The portable NIR spectrometer together with chemometry can be a promising low cost tool to quantify the physicochemical attributes and classify the carambola according to the maturity stage (AU)

• FAPESP's process: 19/12625-1 - Multivariate statistical analyses applied to NIR spectroscopy and digital image analyses for food products
• Grantee: Ingrid Alves de Moraes
• Support Opportunities: Scholarships in Brazil - Master