Analysis of Color Characteristics of Coffee Beans Using Convolutional Neural Networks

Arabica coffee (Coffea arabica) is a type of coffee tree with small beans, which belongs to the Rubiaceae family. Coffea arabica is one of the types of drinks that, when consumed in adequate (established) doses, can have a positive effect on the human body. This is due to the content of key biologically active compounds such as caffeine, lipids, polyphenols, mineral and aromatic compounds. Consumer properties of finished drinks depend on the stage of roasting coffee beans. In the work, a series of 160 digital images of each batch of experimental samples of coffee beans was made: standard roasted, overroasted and underroasted. To determine the optimal stage of roasting coffee beans, the use of artificial neural networks is proposed in the context of applying the “RGB” and “Lab” models, which contribute to a more accurate analysis of the color characteristics of the samples. The results obtained indicate that with an accurate visual album of color images using the “RGB” model for analyzing food systems from the standpoint of quality control, it is possible to monitor the authenticity of food products, including organoleptic indicators.