Authentication of carioca common bean cultivars (Phaseolus vulgaris L.) using digital image processing and chemometric tools.

Abstract Bean authentication can result in higher quality products for commerce. Partial least squares discriminant analysis (PLS-DA) was applied to digital images in order to develop a methodology that allows the non-destructive discrimination of three Phaseolus vulgaris L. cultivars (Agro ANfc9, IPR-Andorinha, and IPR-Sabia) having different technological characteristics. Principal component analysis resulted in a separation of these cultivars, but with a certain amount of overlap. Supervised analysis showed that three PLS1-DA models, each for two cultivars, was moderately better than the simultaneous treatment of all three cultivars (PLS2-DA). Permutation test evaluated statistical significance of PLS-DA models. The classification models were more accurate for Agro ANfc9 and IPR-Sabia cultivars than for IPR-Andorinha. The Agro ANfc9–IPR-Sabia model correctly classified 100% of the two bean classes in both training and test sets. This analytical strategy is fast, inexpensive, environmentally friendly, and can be applied for bean quality control helping cultivar authenticity for commerce.