Identification of Disease in Coffee Leaves Using Local Binary Pattern and Random Forest Methods

Coffee is a very important plantation and one of the largest foreign exchange in Indonesia economy. However, the quantity of coffee production in Indonesia is relatively low. The influencing factors is the death of coffee plants due to infection, so it is necessary to do early prevention by identifying the disease. Identification can be done by looking at changes in color and physical shape of the leaves. Because the coffee plantation area is large with a large number of plants, it is difficult for farmers to identify the disease correctly. In this article, a digital image processing system has been designed that can identify diseases in coffee leaves. Coffee leaf disease was identified into 3 types of disease, Leaf Blight, Leaf Miner, and Leaf Rust. The digital image processing uses Local Binary Pattern and Random Forest methods. The dataset used is 240 images, divided into 192 training data and 48 testing data. From the test results obtained the best accuracy of 95.83% using the parameter image size 128x128 pixels, LBP radius = 1, and n-estimators Random Forest = 100.