Research On Simmental Beef Facial Recongnition Algorithm Based On Deeplearning

The identification of beef cattle is the key to the registration,traceability and monitoring of beef cattle,and it is also the main component of smart pastures.The current identification technology of beef cattle is mainly realized by RFID technology.This method takes a long time,will cause great pain to cattle,and the label is easy to be lost,and the recognition rate is low.Therefore,for Simmental beef cattle with the largest number of domestic cattle farmed in the pasture,this paper proposes a beef cattle identity recognition algorithm based on deep learning and cattle facial features.The specific work is described below.First of all,based on the research of biometrics that can be used for beef cattle identification,a total of 130 videos of 50 beef cattle were collected at the Shengyuan beef cattle breeding base in Hebei,and about 2-3 segments per cow.For a frame of pictures,the facial images are intercepted by selecting key points of the face,and the SSIM image evaluation algorithm is used to remove images with excessive similarity in consecutive frame images to obtain the cattle face dataset of Simmental beef cattle.Then using transfer learning ideas,the VGG-Face network trained on the LFW Face data set recognizes the cow face of the cow face data set designed in this paper,keeping the extracted feature layer(convolutional layer)parameters unchanged,in the fully connected layer Fine-tuning has achieved 91.6% accuracy.Then this article uses the Keras deep learning framework to build a neural network model C_F_R NET(Cow_Face_Recognition NET)for cow face recognition.The impact of network recognition rate was clearly analyzed and explained.Through continuous training and optimization of the neural network,the network has achieved 98.6% accuracy on the cow face data set.Finally,the confusion matrix is used to evaluate the classification model,and the recognition rate of each cow in the data set is obtained.