An Artificial Intelligence and Thermal Imaging Approach for Real-Time Rat Pest Detection in Farming Areas

Abstract

Rat pest attacks in agricultural fields of Sengka Village, particularly at night, cause significant crop damage and economic losses for farmers. Traditional control methods such as traps and manual observation are often ineffective due to limited visibility under low-light conditions. This study aims to develop an AI-based rat pest detection system using thermal cameras capable of operating automatically and in real time. The research methodology includes collecting and augmenting thermal image datasets from Roboflow and Kaggle, training an object detection model using YOLOv11, and evaluating the system through inference on external thermal video data. The results demonstrate excellent performance, achieving mAP@50 above 0.99, precision close to 0.99, and recall exceeding 0.97. The system is able to consistently detect rats and automatically trigger ultrasonic wave emission as a responsive deterrent mechanism upon detection. These findings highlight the strong potential of thermal–AI technology as an early warning and automated pest management solution that can be adopted by farmers, especially in agricultural environments dominated by nocturnal pest activity.