Classification Cherry’s Coffee using k-Nearest Neighbor (KNN) and Artificial Neural Network (ANN)

Abstract

The quality of coffee is determined from 60% when planted, 30% when roasted and 10% when brewed. This research examines more deeply the process of sorting the coffee cherries using the dry method. The technology that is possible to solve this coffee cherry fruit sorting problem is image processing, this is seen because the current conventional method uses human eyes and hands in sorting. This sorting process aims to separate superior fruit (red, half red, broken red, brown), black, half black, orange, yellow, and green) of inferior fruit (spotted, moldy, with 1 hole, and more than 1 hole) and coffee cherries (round, oval, broken, perfect).The purpose of this study was to develop a coffee cherry sorting machine technology with faster and more accurate results so that it could replace the conventional coffee cherry sorting process. The coffee cherries are categorized into ripe, undercooked, raw, and damaged cherries using the GLCM (Gray-Level Co-Occurrence Matrix) algorithm for feature extraction and the KNN (k-Nearest Neighbor) and ANN (Artificial Neural Network) classification algorithm. newrb. The success obtained from this research is ANN accuracy of 24.41% and using the KNN method of 72.12%. With the simulation carried out, the coffee cherries classification process with an amount of 1,885 can be carried out in a total time of 356.02 seconds or the equivalent of 6 minutes. Author identifies indicators of coffee cherries as skin color (red, half red, cracked red, brown, black, half black, orange, yellow, and green), cherri shape (round, oval, broken, perfect), and cherries skin defects (speckled -button, moldy, with 1 hole, and more than 1 hole). It is hoped that the results of this study can serve as a consideration for developing a more advanced national coffee industry.