DESIGN AND CONSTRUCTION OF IOT BASED REAL TIME LIVESTOCK HEALTH MONITORING SYSTEM

Abstract The rapid expansion of the global poultry industry has intensified the need for efficient, accurate, and continuous monitoring of animal health. Traditional manual observation methods are labor-intensive, subjective, and often fail to detect early signs of disease. This project presents the design and construction of an intelligent IoT-based health monitoring system that leverages on-device machine learning (Edge AI) to provide real-time, automated detection of disease indicators in poultry environments. The system utilizes a dual microcontroller architecture, with an ESP32 processing data from a suite of environmental gas sensors (NH3, CO2, H2S, CH4) and an ESP32-CAM for visual monitoring capabilities. A lightweight machine learning model, deployed directly on the ESP32, analyzes ambient gas concentrations to classify the flock's health status as either "Healthy" or "Disease." The machine learning model demonstrated outstanding performance, achieving 98.9% accuracy during training and a perfect 100% accuracy on unseen test data. Crucially, the model is highly optimized for edge deployment, with an inference time of just 1 millisecond and a peak RAM usage of 16 Bytes, enabling instantaneous analysis without reliance on cloud infrastructure. This study concludes that leveraging Edge AI with environmental gas sensing provides a highly accurate, cost-effective, and scalable solution for proactive poultry health management. By enabling early, automated disease detection, the system contributes significantly to the advancement of precision livestock farming, enhancing animal welfare and promoting sustainable agricultural practices.