Computer Vision Methods in the Process of Fermentation of the Cocoa Bean

Abstract

The present work describes different methods or techniques of computer vision, used to estimate the level of fermentation of the cocoa bean. The description of several methods is performed, such as, vectorial quantification, K - means, fuzzy grouping and average displacement. Based on these methods, the training and classification phase of the images has been executed. A simulation platform is used to codify and execute the different methods, obtaining results with a level of certainty of approximately 75%. The images used in this research have been obtained directly during the process of fermentation of cocoa and the reference pattern for the level of fermentation has been developed with information provided by experts in sensory profiles, tasters of cocoa.