Machine Learning Detection of Heat Stress Behaviour in Broiler Chickens

Abstract

Introduction: Heat stress is one of the biggest problems in broiler chicken farming since higher environmental temperatures may have adverse effects on the behavior, health, growth and productivity of birds. Therefore, early detection of heat stress becomes critical for the purpose of ensuring optimal conditions for broiler housing and minimizing production losses. In this study, a machine learning-based method for the detection of heat stress behavior in broilers based on their behavior and visual observations will be considered. For this purpose, videos and environmental data related to broiler housing conditions will be obtained and preprocessed to detect changes in activities such as panting, wing spreading, decreased activity, crowding and changes in feeding/drinking. Image/video pre-processing and feature extraction methods will be used prior to the classification of behaviors via machine learning algorithms. As a result of training the model, a distinction between normal and heat stress behaviors of broiler chickens will be achieved through automated detection of heat stress conditions. Performance of the model will be evaluated through various metrics like accuracy, precision, recall, F1-score and confusion matrix.