Research On Behavior Recognition Of Dairy Goat Based On Multi-model Fusion

The behavior of the dairy goat is an important manifestation of the health of the dairy goat,and the health of the dairy goat will directly affect the economic benefits of the breeder.In order to accurately identify the behavior of the dairy goat in the image,this study took the dairy goat at the Dairy Goat Experimental Station of Northwest A&F University as the object of study,and based on the convolutional neural network,proposed a multi-model fusion-based dairy goat behavior recognition algorithm,And analyzed and compared the recognition results of single model and multiple models,and finally realized the behavior recognition of dairy goats.The main research and conclusions conclude:(1)Construction of dairy goat data set.Firstly,by extracting key frames from the surveillance video of the dairy goat experimental station,deleting invalid images,and analyzing the difficulties of behavior recognition of dairy goat images,a large number of dairy goat images with complex backgrounds are selected,and the initial data set of dairy goat is obtained.Secondly,the data is expanded by the data enhancement method,and finally enough data sets of dairy goats are obtained,which improves the practicability of the algorithm and avoids the over-fitting problem.(2)Research on dairy goat behavior recognition algorithm based on classic convolutional neural network.First,use the classic convolutional neural network models AlexNet,VGG16,Res Net50,Inception-V3,and Inception-V4 models to train on the dairy goat data set to obtain the best recognition results of each model.Secondly,some model has its own improvement method.From the experimental analysis of each model,it can be seen that the recognition effect of VGG16,Res Net50 and Inception-V4 models is the best.(3)Research on milk goat behavior recognition algorithm based on multi-model fusion.In order to distinguish all kinds of behaviors of dairy goats more exactly,and raise the accuracy of the model,a behavior recognition algorithm for dairy goats based on multi-model fusion is proposed.First,select the first 3 models of the recognition results from the above experiments,namely VGG16,Res Net50 and Inception-V4 models.Secondly,the attention weight of each model is calculated through operations such as feature splicing.Finally,the attention weight is fused with the eigenvalue of each model by inner product operation,and the parameters of the fused multi-model convolutional neural network are adjusted to obtain the maximum value of the multi-model.The recognition result is good,and the final recognition result is 98.52%.In summary,the behavior recognition algorithm of dairy goats based on the fusion of multiple models in this article realizes the accurate recognition of various behaviors of dairy goats,It is of great significance to promote the development of current animal healthy.