Application of portable near infrared spectroscopy for classifying and quantifying cocoa bean quality parameters

Abstract

Fermentation duration (FmD), fermentation index (FI), pH, and moisture content (Mc) are vital quality attributes of cocoa beans. In this study, portable near infrared spectroscopy (NIRS) and multivariate analyses were used for rapid determination of FmD, FI, pH, and Mc of cocoa beans. The samples were scanned in 900- to 1,700-nm wavelength, and the spectral data were pretreated independently with first derivatives (FD) and second derivatives (SD), multiplicative scatter correction (MSC), mean centering (MC), and standard normal variate (SNV), while linear discriminant analysis (LDA), support vector machine (SVM), and partial least squares regression (PLS-R) were used to build the prediction models for FmD, FI, pH, and Mc. MSC plus SVM gave an accurate classification of 100%. For predicting FI, pH, and Mc, the PLS-R model gave coefficient of correlation of 0.87, 0.82, and 0.89, respectively. The results showed that portable NIRS could be employed for cocoa bean examination. Novelty impact statement Fermentation is the single most essential postharvest operation that influences cocoa beans quality parameters including moisture content, fermentation index (FI) and pH. Unlike stationary laboratory based wet chemistry technique or table top NIR spectroscopy, this study revealed that the relatively inexpensive portable NIR spectroscopy could provide very fast (within 30 s) results in the routine onsite evaluation of cocoa beans moisture content, fermentation index and pH on farmers field in Sub-Saharan Africa. In particular, the study outcome highlights the potential application of portable NIR spectroscopy based on machine learning for efficient classification of fermentation duration and quantification of moisture content & pH of cocoa beans in real-time usage.