APPLICATION OF DRONE-BASED REMOTE SENSING AND MACHINE LEARNING FOR REAL-TIME HERD HEALTH SURVEILLANCE AND BEHAVIORAL ABNORMALITY DETECTION

Abstract

The integration of drone-based remote sensing with artificial intelligence presents a novel paradigm in real-time livestock health monitoring. This study employed a mixed-method experimental design incorporating both qualitative stakeholder input and quantitative data acquisition using UAVs equipped with multispectral, hyperspectral, and thermal sensors. Over 180 animals were monitored across nine batches, producing 9 detailed tables and 12 complex figures that revealed significant physiological and behavioral patterns. Machine learning models, including convolutional neural networks (CNNs) and support vector machines (SVMs), were trained to detect anomalies using features such as body temperature and motion deviation scores. Results showed that animals with temperatures exceeding 39.2°C and motion deviation scores above 6.5 were consistently identified as high-risk, achieving an anomaly detection accuracy exceeding 90%. Visualizations validated the system’s ability to differentiate between healthy and unhealthy animals in a non-invasive, automated manner. Additionally, a real-time feedback dashboard provided actionable insights to farmers and veterinarians, enabling early intervention and informed decision-making. The proposed system demonstrates strong potential for improving animal welfare, reducing economic losses, and supporting sustainable livestock management practices. This research underscores the viability of smart farming technologies and lays the foundation for future scalable applications in precision animal agriculture.