Mini shortwave spectroscopic techniques and multivariate statistical analysis as a tool for testing intact cocoa beans at farmgate for quality control in Ghana

Abstract

Moisture content and pH are the most important quality parameters measured during the initial processing of cocoa beans at farm gate in developing countries like Ghana. This research investigates the possibility of using portable NIR spectroscopy and multivariate algorithms coupled with smartphone for measuring moisture content and pH at the farm gate. The performances of the model were optimized by cross-validation and examined by correlation coefficient (R) and root mean square error of prediction (RMSEP) in the prediction set. The extracted data were pre-processed with three different techniques (multiplicative scatter correction; MSC, standard normal variant; SNV, second derivative; SD) comparatively and the best was selected. Different types of Partial least square regression; interval Partial least square regression (IPLS), synergy partial least square regression (Si-PLS), and back interval partial least square regression (Bi-PLS) were attempted comparatively to provide the optimum outcome. For the measurements of moisture content and pH in cocoa beans; the best model was Si-PLS with R2 = 0.77 & RMSEP = 3.40 and R2 = 0.73 & RMSEP = 0.44 respectively. This revealed that pocket-sized NIR together with variable selection algorithm (Si-PLS) could be utilized for rapid and non-destructive determination of moisture content and pH in cocoa bean onsite. This will offer a useful tool for cocoa farmers and other quality control officers to ensure cocoa bean integrity in Ghana.