Image clustering using genetic algorithm with tournament selection and uniform crossover

Abstract

This study aims to test genetic algorithm with tournament selection and uniform crossover for clustering complicated images. Genetic algorithm can be used for searching the optimum centroid for clustering images. Images that used in this study is beach images, city images, traditional market images, and garden images. Based on the research result genetic algorithm with tournament selection and uniform crossover can be used for clustering images but there is some outlier in formed cluster. Based on trial the best parameters for image clustering with genetic algorithm with tournament selection and uniform crossover is population=200, iteration=200, and number of cluster=2. Fitness value on genetic algorithm is increase when population and iteration value are higher. The result of this study can be used as a reference in the development of images clustering.