Non-Destructive Prediction of Moisture and Fat Content in Cocoa Beans Using Near-Infrared Spectroscopy and Multivariate Regression Models

Abstract

Presented paper aimed to apply the near infrared spectroscopy (NIRS) for rapid, nondestructive and simultaneous prediction of moisture and fat content on intact cocoa beans. Near infrared spectral data were acquired using a portable NIRS instrument (PSD NIRS i16, Universitas Syiah Kuala) in the wavelength range of 1000-2500 nm. Actual moisture and fat contents were measured and determined by means of thermo-gravimetry and soxhlet methods respectively. Spectral data were enhanced and corrected using standard normal variate (SNV) and de-trending order 2 (Dt-2) methods. Prediction models, used to determine the moisture and fat contents were established using partial least square regression (PLSR) and principal component regression (PCR). The results showed that moisture content and fat content can be predicted non-destructively and simultaneously using NIRS with maximum prediction index (RPD) are 2.16 for moisture content and 2.39 for fat content respectively which can be categorized as good model performances. Moreover, SNV found to be the ebst spectra correction method in predicting both moisture and fat contents.