Study On Individual Recogntion Of Cow Baesd On Gait Feature And Texture

With the increasing attention of the Chinese government to the dairy farming industry,the scale of breeding continues to expand,and the disadvantages of artificial management become more and more obvious.The construction of smart farm has become the development trend of the current breeding industry,in which the individual identification technology has become an important part of the current animal husbandry towards information technology.Combining computer vision technology with individual recognition of dairy cattle can automatically and accurately identify individual dairy cattle,and can improve the level of information construction of farms.In order to improve the robustness,practicability and recognition efficiency of the current cow individual recognition algorithm,this paper proposes a cow individual recognition algorithm combining gait features and texture.The following main work has been done:Firstly,the foreground image in the original cow video is extracted by the foreground detection algorithm,and the original foreground image is processed to generate a gait recognition image.At the same time,the feature extraction algorithms which are widely used in the field of image recognition are described,and three representative classification algorithms are used for experiments.This paper analyzes the advantages and disadvantages of the foreground extraction algorithm,feature extraction algorithm and classification algorithm.Then,aiming at the problem that the recognition rate of cow gait recognition algorithm based on energy graph is not high,and the key feature extraction of cow gait recognition algorithm based on limb angle is less,an algorithm of cow individual recognition based on limb angle energy graph fusion in feature layer is proposed;aiming at the problem that the recognition rate of cow gait is low and the robustness of cow individual recognition based on texture is poor,this paper puts forward a cow individual recognition algorithm based on gait texture fusion in decision level.(1)The feature layer fusion algorithm based on limb angle energy graph is used for gait recognition of dairy cows.Firstly,the gait feature extraction algorithm based on the limb angle is used to extract the two key limb angles of the cow in the process of walking,and the limb angle feature map is generated according to the period;then the improved Local Binary Pattern algorithm is used to extract the local texture features of cow gait energy map,and finally fuses the two features in the feature layer to generate the final feature vector for individual recognition.(2)The decision level fusion algorithm based on gait texture is applied to cow individual recognition.Firstly,the cow foreground image extracted from the original video is preprocessed by normalization and post-processing,and the cow texture is extracted by the improved local binary pattern algorithm.The texture recognition and(1)the cow gait recognition algorithm decision-making layer based on the limb angle energy graph in(1)adopt the method of addition.Finally,using SVM classifier to classify the features of the extracted limb angle energy graph and cow texture features after dimensionality reduction,the recognition accuracy is83.52%,which is higher than the accuracy of single gait based cow individual recognition and single texture based cow individual recognition,which verifies the effectiveness and feasibility of the fusion algorithm proposed in this paper.At the same time,in order to reflect the strong generalization ability of the feature extraction algorithm in this paper,the working feature curves of the subjects drawn by the recognition results of different algorithms on the test set are given.