AI-Driven Automated System for Paddy Disease Detection Using Sensor Networks and Drone-Based Image Analysis

Abstract

Paddy cultivation is vital for Sri Lanka’s food security, but increasing plant diseases due to adverse climate, declining soil health, irregular water availability, and unpredictable weather have caused a continuous drop in yield, highlighting the need for effective disease detection. This study presents an integrated paddy disease detection system that combines Internet of Things based environmental sensing with drone-based remotesensing imagery and artificial intelligence techniques. The proposed system employs an ESP32 microcontroller interfaced with more accurate sensors to monitor soil, water, and agro-climatic parameters in real time. Machine learning models are applied to analyze the collected sensor data and predict potential paddy diseases based on environmental conditions. In parallel, a drone imaging system captures high-resolution images of paddy fields, which are processed using deep learning models developed with Keras and TensorFlow to detect and classify disease symptoms. A Flask-based web application is developed to visualize sensor data, display disease predictions, and provide actionable recommendations for farmers and agricultural officers. Experimental results demonstrate that the proposed system achieves an overall disease detection accuracy of 98%, with additional evaluation using precision, recall, F1-score, and confusion matrix analysis confirming its robustness and reliability. The practicality of the proposed system is enhanced by its low cost, portability, and modular design, enabling easy deployment in small and large paddy fields and allowing scalability to regional and national agricultural monitoring systems.