Research On Template Matching Cow Identification Based On Dairy Activity Dataand CNN

Dairy cow identification is an important part of dairy cow management.Traditional cow identification methods are mainly manual identification,but this kind of work is inefficient and the recognition results vary from person to person.With the development of technology in recent years,electronic tag recognition(RFID)technology and image recognition technology have also been applied to cow identification.However,the ear tag type electronic tag recognition technology is convenient to use,but there is a problem of high drop rate and limited working range,and the rumen type electronic tag has a high cost problem;while the image recognition technology has data collection must work locally,The problem of collecting data to give recognition results cannot be processed in real time.Aiming at the above problems,this paper proposes a method based on cow activity data and template matching of convolutional neural network to identify cows.The method has the advantages of high recognition accuracy and remote real-time cow identification work.This paper mainly studies cows based on cow activity data and convolutional neural network template matching,and proposes cow identification algorithm basedon convolutional neural network and cow identification algorithm based on cow activity quantity eigenvalue,and combines the two to volume The cow identification algorithm of the neural network is used as the preliminary screening algorithm for cow identification.The cow identification algorithm of dairy activity eigenvalue is used as the confirmation algorithm for cow identification.This article has mainly completed the following work:First,the data collection and pre-treatment of dairy activity.Dairy activity data refers to comprehensive information on cow head movement data,swallow volume data,general exercise data,and the like.In this paper,the corresponding data is collected and uploaded to the database for storage by the data acquisition device.The cow activity data entering the database will also be preprocessed by the program to extract the corresponding feature values and store them.The second is to use the convolutional neural network to train and classify the data of dairy activity.Firstly,the cow activity data is made into a waveform diagram of 12 hours period,then the waveform map is trained and classified by convolutional neural network to obtain a classification template,and the tested data is trained by convolutional neural network to obtain a preliminary screening classification.As a result,the result was taken as a preliminary matching result.The third is to propose an algorithm for extracting the characteristic value of the cow activity data and a method for identifying the cow by the feature value.The preliminary matching recognition result is used to identify the eigenvalues of the matching classification group for secondary recognition,and then the confirmationresult is given,and the recognition method effectively improves the accuracy of the cow identification.Finally,the paper also carried out field testing work.Through the research and improvement of the examples,the time length of the required test data is compressed,which further improves the working efficiency of the system,and makes the method have engineering application value.