Autonomous Drone-Based Early Detection and Precision Control of Wheat Diseases

Overview

Wheat disease outbreaks reduce crop yield and make it harder for farmers to maintain stable income, creating a need for faster and more effective detection methods. In many cases, farmers cannot easily identify where a disease begins or how it spreads, which leads them to apply excessive amounts of pesticide across entire fields. This not only increases costs significantly, but also causes serious environmental harm. At the same time, farmers spend thousands of hours manually inspecting individual plants in an attempt to detect early signs of infection. This project develops an autonomous drone-based system for real-time disease detection and targeted treatment in agricultural fields. The drone uses a Raspberry Pi for onboard processing and integrates both a Convolutional Neural Network (CNN) and a Random Forest (RF) model to classify wheat diseases during flight. When a disease is detected, the system is designed to activate a localized spraying mechanism to help prevent further spread. For testing, the spraying system used water to simulate real-world operation without applying chemicals. The RF model achieved an accuracy of 92 percent, while the CNN achieved 90.81 percent. In field testing, the system correctly classified 17 out of 25 grid regions, with an average processing time of 6 seconds per grid. These results show that combining early detection with targeted treatment can reduce unnecessary pesticide use, limit crop loss, and support more sustainable farming practices. This system demonstrates a practical and scalable approach to improving crop health management.