KNN FOR CLASSIFICATION OF FRUIT TYPES BASED ON FRUIT FEATURES

Abstract

Research related to the recognition of fruit types has been done previously. Research related to the recognition of many types of fruit applies computer vision and artificial intelligence. The purpose of this research is to apply artificial intelligence science with the KNN method to identify the type of fruit. The KNN method has a good performance in previous studies. We tried to use KNN by determining the most optimal K value. There are five types of fruit images used in this study, namely Apples, Grapes, Oranges, Mangoes, and Strawberries. The fruit image is extracted with colour, texture, and shape features with a total of 15 features, namely the average value of R, the average value of G, the average value of B, the value of skewness R, the value of skewness G, the value of skewness B, the value of grayscale entropy. , grayscale contrast value, grayscale energy value, grayscale correlation value, grayscale homogeneity value, binary area value, binary circumference value, binary major axis value, and binary minor axis value. The dataset used in this study was taken from Kaggle, with a dataset of 2750 images, each type of fruit contained 550 images, 2500 training images were used and 250 images were used for testing. The experimental results show that the KNN method with K=1 has the highest accuracy, which is 99.6%. The KNN method can be used optimally in classifying fruit types based on colour, texture, and shape features.