A Smart Agriculture Framework to Automatically Track the Spread of Plant Diseases Using Mask Region-Based Convolutional Neural Network
Abstract
Plant diseases reduce agricultural production. They negatively affect fruit and crop quality and reduce yield, causing food shortages. A drop in production harms the global agricultural economy. However, early detection and disease severity estimation are key to disease management, containment, and prevention. Damage localization is the first step in estimating the severity of diseases, which is crucial for the optimum application of pesticides. The current approach needs expert advice for disease detection. For a large farm, it is an expensive and slow process. Automatic plant disease detection eliminates the tiresome task of monitoring big farms and detects the disease early enough to avoid plant degradation. In this article, we propose a fully automated method based on deep neural networks for detecting and localizing leaf diseases. The proposed method is based on Mask R-CNN network. Image augmentation has been performed to achieve higher precision from a small dataset. Transfer learning has been used to save time and achieve better performance. Our proposed method of disease detection is faster, as it automatically localizes the disease along with the disease identification from the leaf images. The images can be taken using a smart phone camera or a low altitude unmanned aerial vehicle (UAV) camera. As a case study, we have applied the method to apple leaves.
Origin | Files produced by the author(s) |
---|