畜禽个体识别技术研究进展OA

Individual identification of livestock and poultry was an important prerequisite for achieving refined management and intelligent breeding.Traditional manual identification methods such as earprint ear patterning and hot iron branding faced issues of low identifica-tion efficiency and high individual stress.Individual identification based on radio frequency identification technology methods offers low stress levels but came with drawbacks such as high costs,prone to detachment and short battery life.With the rapid development of machine vision and deep learning technology,non-contact individual recognition methods had become one of the current research hotspots.On the basis of thoroughly reviewing the existing individual recognition methods,this paper introduced typical contact individ-ual recognition methods and their advantages and disadvantages,described two non-contact individual recognition methods based on image processing and deep learning and their advantages and disadvantages.Summarized and analyzed the problems and improve-ment strategies of deep learning in animal individual recognition,including deep learning models,sample data size and research level.Relevant suggestions were proposed,which could provide theoretical basis and technical support for breeding management personnel.