Research And Application Of Pig Video Tracking And Monitoring System Based On Machine Vision

With the rapid development of Internet technology,machine vision technology and other modern technologies,intelligent monitoring system has been promoted and applied in agricultural production and breeding,and has become an essential part of intelligent breeding.Intelligent and healthy pig breeding is related to people's life health.In this paper,the application of intelligent monitoring system in pig house environment can help to reduce human and animal contact,improve breeding efficiency,and its automatic recognition and alarm function can make up for the problem that human can't achieve 100% concentration in work,form the complementary advantages and disadvantages of human and machine vision,and meet the requirements of current pig breeding.In order to realize this system,the main contents of this paper are as follows:First of all,a pig video acquisition system is established on site,including the selection of camera,the selection of installation location,the construction of server and database,and the access of video image.Then,the pig house monitoring video image is used as the data source to complete the research of pig recognition and tracking algorithm,In the aspect of recognition,after comparing the traditional and deep learning target detection algorithms,a pig target detection method based on the improved SSD(single shot multibox detector)algorithm is proposed.Based on the DSOD(Deeply Supervised Object Detector)algorithm and the actual pig target,the improved idea solves the problems of less training sets and less pig target feature selection;In the aspect of tracking,it is found that a single tracking method can't meet the requirements of the system through the experiments of background difference method,optical flow method and other tracking algorithms.Therefore,a pig video tracking algorithm based on the improved TLD(tracking learning detection)framework is proposed.The improved algorithm retains the integrated module and p-n learning mechanism of the original TLD framework,and replaces the detection module and tracking mechanism of the original TLD algorithm Module improves algorithm performance.Finally,the relevant codes are written on Python platform to realize the functions of monitoring video reading,pig target detection,live pig video tracking,etc.,and the user interface is written on this basis.The user interface functions include the display of monitoring video,the drawing of tracking track,the query of historical data and the alarm of abnormal conditions.The staff log in the user interface to realize the above functions Now human-computer interaction.In this paper,on the basis of drawing the daily action track of pigs,the daily rest and activity rules of pigs are analyzed,and the daily activity evaluation table and health status analysis table of pigs are established,which are used to automatically detect the health status of pigs.The system will provide alarm through the user interface when finding the abnormal situation,remind the staff to check and handle in time,which lays the foundation for the realization of the comprehensive automatic intelligent monitoring of the farm,and has a good application prospect.