Unmanned Aerial Vehicle-Based Large Area Disease Detection of Paddy Fields using Machine Learning

Early disease detection in agriculture is critical to ensuring sustainable agricultural practices and maintaining high yields. Diseases cause significant economic losses and food insecurity. Delayed detection often leads to the extensive spread of pathogens, making recovery challenging and resource-intensive. Although current practices, such as hyperspectral and multispectral imaging systems, provide precise disease detection through advanced spectral analysis, their high costs, complex data requirements, and limited accessibility restrict their adoption by small to medium-scale farmers. Addressing these challenges, this research develops a cost-effective UAV-based remote sensing system leveraging a low-cost method with RGB cameras and machine learning techniques for early disease detection in large paddy fields. To enhance detection accuracy, the study has utilized vegetative indices such as VARI, GLI, ExG, and VIgreen. Random Forest machine learning algorithm outperforms Support Vector Machine, K-Nearest Neighbour, and Logistic Regression. Random Forest processes these vegetative indices to classify diseased areas with 68% of accuracy and 73% of precision. Field testing validated the system’s performance, demonstrating its ability to deliver reliable results. This scalable, low-cost approach bridges the gap between advanced imaging technologies and practical agricultural needs, offering an accessible solution for efficient crop health monitoring.