Design And System Implementation Of Target Detection And Sample Data Automatic Extraction Algorithms Based On Video Cow's Face

With the rapid development of computer application technology,the use of intelligent and modern technology to manage livestock farming enterprises has changed the traditional management mode for an enterprise,bringing great benefits to both human and financial resources.Based on convolution neural network,this paper studies the automatic detection algorithm of dairy cows and its system implementation,and realizes the real-time target detection task of dairy cows in milking environment efficiently and accurately.At present,because there is no standard data set in the study of individual identification of dairy cows,at present,all open projects produce data sets for experiments by pure manual method.Therefore,one of the purposes of this paper is to realize the automatic collection of sample data sets.In addition,the automatic collection of sample data sets can greatly improve the automation level of target detection and recognition system,and further improve the efficiency of project implementation in the process of system integration and demonstration promotion.The main contents of this paper are as follows:(1)Video capture of dairy cows.In order to study the target detection effect of the algorithm for dairy cow video,field video image acquisition of dairy cow was carried out.In the process of video acquisition,in order to avoid unnecessary troubles to the daily production of dairy farms as much as possible,and because of the complexity and uncertainty of the environment of dairy farms,a scheme of how to obtain the video image of dairy cows was developed,and the front-face video of dairy cows passing through the turntable in a single background was collected.(2)Technical scheme.The advantages and disadvantages of the traditional target detection algorithm and the candidate region-based target detection algorithm and the deep learning-based regression target detection algorithm are compared.Finally,the deep learning-based regression target detection algorithm is used as the research topic.And based on the deep learning network model SSD,the improved VGG16 is used as the framework of data training.Since there is no standard data set in the research of dairy cow target detection,the data set used in experiments is entirely hand-made by ourselves.The experimental results show that the scheme can effectively detect the real-time target of dairy cows in milking table environment.(3)System design and implementation.On the basis of algorithm research,this paper takes Windows 10 as the platform and PyQt5 as the platform to independently develop a graphical interface,which can display the results of the current target detection of dairy cows in real time,and design and implement the technology of automatically extracting the detected target and generating sample data sets.The interface of the system can transform the related algorithms into visual graphical operation interface.The test shows that the graphical interface is very practical.