Design And Implementation Of Abnormal Detection System For Broiler Legs Based On Infrared Thermal Imaging Technology

With the rapid development of broiler industry in China,the level of intensive and large-scale breeding has been continuously improved,while in the process of production,broiler leg disease occurs frequently due to excessive breeding density,improper management,epidemic situation and other factors,which restricts the daily activities and production benefits of broilers.At present,the early diagnosis of broiler leg disease is mainly based on human eye observation,which is time-consuming and subjective.The use of modern technical means to realize the automatic detection of broiler leg disease is helpful to improve the welfare breeding level of broilers and reduce the economic loss caused by the disease.According to the needs of broiler breeding and production,this paper takes white feather broiler as the research object,and designs a broiler leg anomaly recognition and early warning system by using infrared thermal imaging technology,sensor technology,neural network model and so on.The integrated algorithm model of the system can be used for data mining,automatically detect and store abnormal information of broiler legs,and realize data visualization management through WEB platform.The main research contents are as follows:(1)Design of hardware system for abnormal detection of broiler legs:Haekangwei TB-1217A-3/PA thermal imaging camera is selected to collect infrared thermal image and temperature data,and an environmental information perception system of chicken house based on NB-Io T technology is designed to realize real-time monitoring of ambient temperature,relative humidity and light intensity in the chicken house,and upload the data to the cloud server through the gateway.(2)Study on the inversion model of leg temperature of broiler: using YOLOv4 target detection algorithm to identify the region of interest((Region of Interest,ROI),by extracting ROI temperature and combining environmental factors,the inversion model of leg temperature of broiler was established,and the multiple linear regression and KNN regression were compared.The average relative errors were 0.71% and0.43%,respectively.Finally,KNN regression was selected as the temperature prediction model.(3)Research on classification and recognition model of abnormal legs of broilers: the infrared thermal image was binarized by OSTU adaptive threshold segmentation algorithm.By extracting posture features and combining temperature features,a classification and recognition model of abnormal legs of broilers based on random forest(Random Forest,RF)was established.The recognition accuracy of the model was 97%,91% and 94% in normal,slightly abnormal and moderately abnormal levels,respectively.The overall accuracy is 96%.Through the comparison of different models,the RF model performs best.(4)Realization of application system for abnormal detection of broiler legs:according to the functional requirements of the system,My SQL database is used to store and manage environmental monitoring information,abnormal information of broiler legs and administrator information.Middleware developed based on Java can monitor,read and process multiple folders.WEB background service is developed based on lightweight framework Spring Boot,and WEB front end is designed by using Java Script,HTML and other technologies.Through the introduction of Highcharts,Echarts and other chart plug-ins to achieve a variety of data display,the test results show that the system has a good load-bearing capacity and can run stably in the production environment.