Abstract

PORTILLO MENDOZA, Pedro Miguel  and  PONCE ALVINO, Jefferson Peter. Optimum classification of coffee fruits according to their maturity by means of a control algorithm. Rev. Inv. Cs. Agro. y Vet. [online]. 2022, vol.6, n.18, pp.441-452.  Epub Nov 04, 2022. ISSN 2664-0902.  https://doi.org/10.33996/revistaalfa.v6i18.181.

The purpose of this research is to know to what extent an algorithm-controlled automatic system allows the optimal classification of coffee fruits according to the degree of maturity, identifying them by their color. For which a multilayer neural network was developed using MATLAB which was implemented in a STM32F103C8 microcontroller, using as input data the RGB color mode characteristics of 300 samples of coffee fruits in different stages of maturation, delivered by a sensor of color TCS3200, which allowed having a database of different maturity levels used to train the multilayer type neural network with 3 inputs; 3 hidden layers with 6 neurons in the first layer and 3 in the other two, as well as one neuron in the output layer. The data was organized according to the state of maturity of the fruits, in "Optimal Maturity" or "Non-Optimal Maturity". The system was tested with 60 coffee fruits, obtaining as a result an efficiency of 96.67% and an error rate of 3.33%; thus confirming that the classification system through the control of the algorithm and multilayer neural network designed, identifies and classifies based on the maturity of the coffee fruits optimally.

Keywords : Coffee classification; Algorithm; RGB colors; Neural network; Control algorithm.

        · abstract in Spanish | Portuguese     · text in Spanish     · Spanish ( pdf )