Citrus Leaf Variety Identification Based on RGB and HSV Color Features

Indonesia has a large variety of agricultural production. One type of agricultural production is citrus fruit. Several studies have developed ways to identify types of citruses from the leaves. One of them is classifying types of orange leaves from LBP (Local Binary Pattern) and moment invariance features. Based on previous research, the classification process only calculates the closest distance to the orange leaf feature. The results of measuring similarity based on distance have a high level of accuracy. Based on this, this research propose a classification of orange leaf types from RGB and HSV colors using the Machine Learning method (k-NN (k-Nearest Neighbor), SVM (Support Vector Machine), and Decision Tree). The aim of this research is to determine the accuracy results of the features and methods for classifying types of orange leaves. This research collected data by photographing orange leaves in the garden. The dataset used is 120 leaf images, consisting of 60 lemon and 60 nipis. Features taken from RGB (Red, Green, Blue) and HSV (Hue, Saturation, Value) image leaves are mean, standard deviation, skewness, entropy, variance and kurtosis. The methods for classifying citrus leaf types are k-NN, SVM and Decision Tree. The best accuracy result is the k-NN method, with an accuracy of 0.975.