Exploration of machine learning approaches for automated crop disease detection

Abstract

In the era of frequently changing climatic conditions along with ever increasing world population, it becomes imperative to ensure food security. The burden of biotic stresses pose serious threat to crop productivity, therefore, early and accurate detection of plant diseases is essential. Conventional methods exclusively rely on human expertise, and are often labor-intensive, time-consuming, and prone to errors. Recent advancements in machine learning (ML) offer promising alternatives by automating the disease detection processes with high precision and efficiency. We comprehensively analyze various ML techniques, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Support Vector Machines (SVMs), Random Forest (RF), and Deep Learning Architectures like ResNet and Inception, among others, highlighting their methodologies, datasets, performance metrics, and real-world applications. This systematic review provides a comprehensive analysis after text mining the most recent literature resources of the last half a decade. The review discusses the proposed models, techniques, accuracy, feature selection, extraction methods, the types of datasets used to perform experiments, and the sources of the datasets. Additionally, this review provides critical analyses of existing models in the context of their limitations and gaps. Our findings suggest that while ML based methods demonstrate substantial potential for enhancing agricultural disease management, there is a urgent need for more robust, scalable, and adaptable solutions to address diverse agricultural conditions and disease complexities. By systematically analyzing the extracted data, this review aspires to provide a valuable resource for researchers and practitioners aiming to develop and implement ML-based systems for crop disease detection, thereby contributing to sustainable agriculture and enhancing food security.