Orange Leaf Diseases Identification Using Digital Image Processing

Fruit cultivation is one of the major cash crops in China.Fruit production in China has increased since the early ?90s.Citrus to be particular has increased in production since 2014,China being the second producer of citrus fruits in the world.Recently orange production has dropped in the result of the continuing impacts of citrus diseases such as citrus greening disease.There have been several ways to detect the diseases which includes diagnosing or sometimes guessing about the disease by using experience which is mostly done visually by humans which come with errors and wastage of time but with the advancements of digital technology,it becomes easy to identify plant disease with image processing techniques.This thesis proposes a method to identify disease symptoms in navel oranges? leaves using images obtained in real conditions of the field.The approach proposed consists of the following steps;image segmentation,feature extraction followed by training and classification.Experiments on color spaces of images were carried out to obtain the best threshold for segmenting diseased parts of the leaf.Color,texture,and local features are used for training multiple classifiers and finally,classification is done to detect the disease if present.The approach is then tested on the identification of two common diseases affecting the navel oranges of Jiangxi;Citrus Greening and Citrus Canker and the accuracy achieved was 91% by Random Forest and K-Neural Network classifiers.