基于改进MobileNetV2的牛只面部识别方法研究OA

In view of the problems that cattle are easily frightened and consume manpower and material resources in the process of traditional cattle individual identification,this paper proposes a non-contact cattle individual identification method by means of computer vision technology.Firstly,the cattle face photos are taken in the real cattle farm environment,and the cattle face recognition dataset is constructed.Secondly,combined with the actual network conditions and computing power level of the cattle farm,the lightweight Neural Network model MobileNetV2 is selected as the basic network model.Finally,the Coordinate Attention Mechanism is introduced to improve the MobileNetV2 model,so as to improve the accuracy of the model and enhance its feature extraction ability for key positions.The experimental results show that the improved MobileNetV2 model performs well in the cattle face recognition task.The model size is only 2.86 MB,and the recognition accuracy is as high as 95.81%,which can fully meet the needs of non-contact cattle individual recognition in the actual cattle farm environment.