Recent advances in image processing techniques for automated leaf pest and disease recognition – A review

Fast and accurate plant disease detection is critical to increasing agricultural productivity in a sustainable way. Traditionally, human experts have been relied upon to diagnose anomalies in plants caused by diseases, pests, nutritional deficiencies or extreme weather. However, this is expensive, time consuming and in some cases impractical. To counter these challenges, research into the use of image processing techniques for plant disease recognition has become a hot research topic. In this paper, we provide a comprehensive review of recent studies carried out in the area of crop pest and disease recognition using image processing and machine learning techniques. We hope that this work will be a valuable resource for researchers in this area of crop pest and disease recognition using image processing techniques. In particular, we concentrate on the use of RGB images owing to the low cost and high availability of digital RGB cameras. We report that recent efforts have focused on the use of deep learning instead of training shallow classifiers using hand-crafted features. Researchers have reported high recognition accuracies on particular datasets but in many cases, the performance of those systems deteriorated significantly when tested on different datasets or in field conditions. Nevertheless, progress made so far has been encouraging. Experimental results showing the leaf disease recognition performance of ten CNN architectures in terms of recognition accuracy, recall, precision, specificity, F1-score, training duration and storage requirements are also presented. Subsequently, recommendations are made on the most suitable architectures to deploy in conventional as well as mobile/embedded computing environments. We also discuss some of the unresolved challenges that need to be addressed in order to develop practical automatic plant disease recognition systems for use in field conditions.