Identification of Citrus Canker on Citrus Leaves and Fruit Surfaces in the Grove Using Deep Learning Neural Networks

Abstract

Citrus canker (Xanthomonas axonopodis pv. citri) is a bacterial disease of citrus tree leaves and is the most feared citrus disease, affecting all types of important citrus crops. Currently, there is no cure for citrus canker but experimental prevention methods. Citrus canker is a quarantine disease, meaning that citrus fruits with citrus canker cannot be transported between countries or markets, and this is a huge threat to the citrus production industry. Citrus exports to international markets are highly restricted today due to fruit diseases such as citrus canker because fruits are quarantined for bigger world markets such as European Union (EU) and United States of America citrus markets. For a row-guided robot to detect citrus canker infections on leaves and fruits in the grove, there is a need for an accurate method that can detect defects on the leaves and surface of fruits, classify citrus leaf and fruit images on-the-go, and discriminate citrus canker from other defects. Previous methods in machine vision could detect defects on the surface of citrus fruits, though, the methods were unable to differentiate defects. The neural network system invents edge detection, and automatically learns to characterize the image in terms of edges that appear in the image, and give a more succinct, higher-level representation than raw pixels. A standard strategy in deep learning neural networks is to run the learning algorithm with many optimization parameters and pick the model that gives the best performance on a validation set. The authors are optimistic that identification of citrus canker on citrus leaves and fruit surfaces in the grove could be one of the agricultural problems that can potentially benefit from deep learning neural network approach, and consequently, help to eradicate the predicaments facing the citrus industry.