Design And Implementation Of Diseased Chickens Detection And Henhouse Environment Monitoring Platform Based On YOLOv5s

The domestic broiler breeding industry is developing in the direction of scale,standardization and mechanization.Epidemic prevention and control and environmental monitoring of chicken houses are important tasks in the process of intensive breeding.Poor house conditions can adversely affect flock health and reduce productivity.To prevent the occurrence of diseases in chicken flocks,the environmental monitoring of chicken houses and the technical means of detection of sick chickens should be improved.At present,the prevention and control of epidemic diseases in chicken farms relies heavily on manual identification.Relying on manual experience to judge has limitations,and it is impossible to identify sick chickens in time.The image recognition technology can identify the collected images,determine whether there are sick chickens in the video or pictures,so as to detect the sick chickens in time and give an early warning,so that the staff can quickly deal with the sick chickens and avoid missing the best time for processing.At the same time,it reduces the harm of the source of disease,controls the source of infection,and reduces the risk of other healthy chickens getting sick,thereby reducing the economic loss of farmers.In this regard,the main research contents of this paper are as follows:(1)Construction of sick chicken identification model based on YOLOv5s-CBAM algorithm.First,the dataset was collected and preprocessed,and more than 4,000 images were classified and annotated.By comparing the experimental results of the YOLOv3,YOLOv3-Tiny,and YOLOv5s algorithms,it is estimated that the YOLOv5 algorithm has the best overall performance.In order to improve the performance of the diseased chicken detection model,this paper adds two attention mechanisms,SE and CBAM,on the basis of the original YOLOv5s.The experimental data show that the YOLOv5s-SENet and YOLOv5s-CBAM model algorithms both improve the performance as a whole.Among them,YOLOv5s-CBAM performed the best,with an increase of 1%,0.4% and 0.6% in terms of accuracy,mAP＿0.5 and mAP＿0.5: 0.95,and the weight file was reduced by 3.7mb,which is about 23% of the original model.A single image is 16% faster.(2)Build a functional module for environmental monitoring of chicken houses.Environmental factors in the chicken house will affect the growth and development of chickens,mainly including light,temperature and humidity,and harmful gases.Through a variety of sensors,the changes of the chicken house environment can be understood in real time,and the control equipment can be used to adjust the chicken house environment appropriately to provide a suitable growth environment for the chickens.(3)Developed a set of sick chicken detection and chicken house environment detection system.The front-end core framework of the system is Vue,the back-end core framework is Spring Boot,and MySQL is used for data persistent storage.The system integrates four functions of sick chicken identification,chicken house environment monitoring,data display,and authority management.Users can use this system to realize real-time control of the chicken farm status.