Analysis of a cans waste classification system based on the CMYK color model using different metric distances on the k-means method

Abstract

Abstract This study aims to build and analyze a classification system of can waste based on Cyan, Magenta, Yellow, and Black (CMYK) digital image color model by implementing 3 different metric distances on the k-means method; Manhattan, Euclidean, and Minkowski. The classification results of experimental data note that the implementation of Euclidean distance on the k-means clustering method for classifying the cans waste into three can types has the highest accuracy, with a difference of not more than 1% from Minkowski distance and more than 19.6% from Manhattan distance. The simulation study of various size of generated data show that classification accuracy level using the three metric distances for both the lighting data have a rate less than 70%. The accuracy level of less than 70% in both experimental and simulation data, each of which implements three distances, can be said that this method is not appropriate for building a can classification system.