Identify fungal diseases of cucumber (Powdery Mildew and Anthracnose) using image processing and artificial neural network approach

Abstract

Plant disease can cause reduce quality and quantity of agriculture crops. In some countries farmers spend considerable time to consult with plant protection, while the time is an important factor to controlling of disease. Due to the fact that Powdery Mildew and Anthracnose fungal diseases cause the most damage in cucumber greenhouses, in this study, by presenting a non-destructive method based on image processing technique and artificial neural network, these two types of fungal diseases have been diagnosed. The steps related to the implementation of the proposed method are divided into three parts: segmentation, separation of damaged parts from the leaf and classification of the disease type class. After color and texture features were extracted from cucumber leaf samples, a multilayer perceptron neural network with error post-diffusion learning algorithm was used to separate different classes of images. Network input is the average of the main color components (R, G, B) of the images and the output is zero as a healthy leaf, number one as Powdery Mildew and number two as Anthracnose. The structure of this network was 24-3-4-3, which uses the tansig transfer function for the hidden and output layer, and among the educational functions. So back propagation (BP) algorithm in neural network by using lovenerg marquart (LM) function training has been successfully to diagnosis and classifies plant diseases in 6 second with 99.95% accuracy.