Tracking guided actions recognition for cows

Abstract

BackgroundCows actions are important factors of cows health and their well‐being. By monitoring the individual cows actions, we prevent cows diseases and realize modern precision cows rearing. However, traditional cows actions monitoring is usually conducted through video recording or direct visual observation, which is time‐consuming and laborious, and often lead to misjudgement due to the subjective consciousness or negligence.MethodsThis paper proposes a method of cows actions recognition based on tracked trajectories to automatically recognize and evaluate the actions of cows. First, we construct a dataset including 60 videos to describe the popular actions existing in the daily life of cows, providing the basic data for designing our actions recognition method. Second, eight famous trackers are used to track and obtain temporal and spatial information of targets. Third, after studying and analysing the tracked trajectories of different actions about cows, a rigorous and effective constraint method is designed to realize actions recognition by us.ResultsMany experiments demonstrate that our method of actions recognition performs favourably in detecting the actions of cows, and the proposed dataset basically satisfies the actions evaluation for farmers.ConclusionThe proposed tracking guided actions recognition provides a feasible way to maintain and promote cows health and welfare.