Research On Cattle Daily Behavior Recognition Based On Computer Vision

With the large-scale and intensive development of China’s breeding industry,the behavior supervision of livestock has become particularly important.It can provide data resources for disease prevention,rational feeding and breeding,and help to pay better attention to livestock welfare and breeding efficiency.However,at present,China’s animal husbandry is still dominated by artificial breeding.This traditional breeding method generally has some problems,such as backward infrastructure,extensive management,small scale,low efficiency and poor efficiency.With the proposal of the concept of intelligent breeding,livestock behavior recognition based on sensors is booming,but this method needs to wear sensors for livestock,which is easy to cause stress response and inconvenience to the normal activities of livestock.In this paper,a recognition method of cattle daily behavior based on computer vision is proposed,which realizes the non-contact and automatic real-time monitoring of cattle daily behavior.The research work of this paper mainly includes the following contents:Firstly,the daily behavior data set of single target cattle in the farm is constructed,the daily behavior of single target cattle is identified by yolov5 s target detection algorithm,and the daily behavior time statistics algorithm of single target cattle is designed to realize the statistics of daily behavior time of single target cattle,so as to provide reference value for subsequent multi-target cattle daily behavior identification.Secondly,the daily behavior data set of multi-target cattle in the farm is constructed,the daily behavior of multi-target cattle is identified by yolov5 x target detection algorithm,and the statistical algorithm of daily behavior time of multi-target cattle is designed to realize the statistics of daily behavior time of multi-target cattle.Next,the Yolov5-deepsort target tracking algorithm is used to track the ID of cattle and identify the daily behavior of cattle.Finally,an algorithm is designed to calculate the daily behavior time of every cow under the ID of each cow by combining with the Yolov5-deepsort target tracking Algorithm,and the ID tracking of the cattle in the farm is realized,the identification of cattle’s daily behavior and the statistics of the occurrence time of cattle’s daily behavior under each ID provide the data basis and a technical solution for the daily breeding management and the analysis and evaluation of cattle’s health status,provide technical support for intelligent farming and precision farming.