AgriDrone AI: An Intelligent UAV-Based System for Real-Time Plant Disease Detection Using YOLOv8 and IoT Integration

Abstract

Agriculture faces significant challenges from plant diseases, resulting in substantial economic losses and reduced crop yields. This paper presents AgriDrone AI, an intelligent system integrating Unmanned Aerial Vehicles (UAVs), Internet of Things (IoT) sensors, and deep learning for real-time plant disease detection. The system has been developed and validated as a bench-top payload, demonstrating end-to-end functionality including image acquisition, processing, disease detection, and data visualization. While the system is designed for UAV integration and all components have been bench-validated, full in-flight validation remains as future work. The system employs an ESP32-CAM for aerial image acquisition, an ESP32 microcontroller with DHT11 sensor for environmental monitoring, and GPS for precise geolocation. Images are transmitted via HTTPS to a ground server where OpenCV preprocessing enhances image quality before YOLOv8 inference. The model was trained on the PlantVillage dataset containing 25,274 images across 29 disease classes. Experimental results demonstrate robust performance with precision of 0.7426, recall of 0.6722, and mAP@50 of 0.7319. The system achieves an inference time of 243.75 ms and processing rate of 4.07 FPS, confirming suitability for real-time agricultural applications. The web-based dashboard provides intuitive visualization of disease detection, environmental metrics, and geospatial mapping, enabling informed decision-making for precision agriculture.