Research On Sheep Face Identity Recognition Based On Improved Deep Convolutional Neural Network

As a major livestock region in China,the livestock industry represented by sheep farming is one of its pillar industries and occupies an important proportion in the national economy,which is both a basic industry and a characteristic industry.At present,the introduction of foreign livestock diseases,weak infrastructure of disease control system and the continuous occurrence of livestock drug abuse have brought great pressure and challenges to the traceability of cattle and sheep diseases and the supervision of livestock product quality and safety.With the development of society,China’s animal husbandry industry will change from traditional animal husbandry to modern animal husbandry,and the trend of large-scale breeding will become a trend,and the application of modern equipment and technology will be expanded,so intelligent animal husbandry is an important development direction for the development of national animal husbandry welfare breeding,of which livestock identification is the key and foundation of intelligent farm management.With the rapid development of deep learning and machine vision technology,the wide application of individual livestock identification has been realized.However,due to the inbreeding phenomenon of sheep,their similarity in looks is high and difficult to identify,which makes it difficult to build a sheep identification system in sheep farms.In order to solve the above problems,this paper focuses on sheep face identification based on deep convolutional neural network,and the specific work is as follows:1.Improved YOLOv5 based attention mechanism for sheep face facial region image acquisition.To achieve the acquisition of facial region images of sheep faces,this paper uses the YOLOv5 model as the basis of the sheep face facial region image acquisition model,and adds the SE attention mechanism network model to the C3 module in YOLOv5 to form the C3 SE module,and also adds the depth separable network to improve its operation speed.The experimental results show that the average accuracy of the improved sheep face facial region image acquisition model reaches 99.4%.It provides a high-quality dataset for the subsequent sheep face recognition.2.Sheep face recognition model based on attention mechanism of convolutional neural network.In order to solve the problem of difficult recognition between similar individuals of inbred sheep,this paper uses the traditional convolutional neural network VGG16,Alex Net and Google Net as the basic sheep face recognition model.In addition,an attention network model and a depth-separable convolutional model are added to the traditional convolutional neural network to extract features of sheep face features at different recognition scales,and to improve the network training speed and the extraction accuracy of sheep face features at the same time.The experimental results show that the improved Google Net model has the best recognition rate of 97.16% among the six experimental models.3.A sheep face recognition model based on asymmetric bilinear convolutional neural network.To solve the problems of lower recognition rate and larger models of sheep face recognition model based on attention mechanism convolutional neural network,this paper proposes a sheep face recognition model using VGG19 and Res Net50 as benchmarks,removing its fully connected layer and classification layer and only retaining its feature extractor to extract different features of the image respectively,after which the features are fused together to form a bilinear feature vector of this image for recognition.The experimental results show that the recognition accuracy of the improved model reaches99.69%,which is 2.53% higher than that of the improved Google Net model,and its model size is only 1/3 of that of the improved Google Net model,achieving a lightweight sheep face recognition model with improved accuracy at the same time.4.Design and implementation of a sheep face recognition system based on Py Qt5.In order to realize the visualization and usability of sheep facial identity recognition,this paper develops an identity recognition system based on deep learning of sheep facial features based on the sheep face recognition model of asymmetric bilinear convolutional neural network.Finally,the real-time detection and recognition of sheep faces is achieved and the recognition results are visualised and displayed.