Empirical Study on Ewe Estrus Behavior Detection Based on Target Recognition and Tracking Technology
With the industrial application of digital intelligent monitoring technology in livestock farming,how to utilize deep machine learning techniques to deeply mine and apply monitoring video data from livestock farming,to upgrade from the current rough and low-precision identification to more precise high-precision analysis,and to delve from intuitive shallow features into complex associated deep features,has become a critical technical challenge that the current livestock environmental monitoring system urgently needs to be addressed.At present,the detection of estrus behavior in ewes mainly relies on manual observation and special wearable devices,which has the problems of large error,easy to cause stress reaction and high cost in ewes.This paper focuses on the distinctive features of estrus in ewes,which include frequent walking and decreased appetite,and selects a sheep farm in Xiaolanwo Village,Tai'an City,Shandong Province,as the experimental site,based on real-time video data recorded by the digital livestock monitoring system.Firstly,the Yolov5 object detection model is used to detect and identify estrus ewes;then,the DeepSORT target tracking algorithm is employed to obtain the real-time location of estrus ewes during feeding activities.By extracting the target's position coordinate data,Finally,a walking distance detection model and a walking main trajectory trace detection model during feeding are designed to achieve real-time and accurate detection of estrus ewes.The research provides further theoretical exploration and feasible application solutions for a reasonable grouping and precise and efficient breeding management of small groups of ewes under intensive confinement.
Deep learninglivestock breedingewesestrustarget tracking
国家科技期刊平台
NSTL
万方数据
