Study On Health Monitoring Of Broiler Flocks And Removal Test Of Dead Broilers

Large scale cage broiler farms are prone to the production of diseased and dead broilers.If not detected and cleaned up in a timely manner,it can easily breed pathogens,seriously affecting the quality and yield of broilers.At present,broiler farms mainly rely on manual inspection to detect dead broilers and carry out manual removal,which is labor-intensive and easy to cause dead broilers to stay due to visual fatigue;When the breeder picks up dead broilers,it is easy to bring in external pathogens,which can bring stress and disease risks to surrounding healthy broilers.Therefore,it is of great significance to study the health monitoring of broiler flocks and the removal devices for dead broilers in response to the problems existing in large-scale cage raising of broiler broilers.The main content of this study is as follows:(1)Design and construction of a broiler flock image acquisition device.This study investigated the breeding site of broiler broilers and built an image acquisition device based on the stacked cage mode of broiler broilers.The device uses a wheeled chassis to automatically collect images of the cage broiler population,and the collected image data is transmitted in real-time to the upper computer located on the same LAN through a wireless transmission module.The upper computer monitors the health status of the broiler population based on behavior recognition and thermal infrared technology,and identifies dead individuals.(2)Research on the health monitoring method for broiler flocks based on behavior recognition.By comparing the detection performance of different deep learning algorithms on visible light image datasets of broiler broilers,YOLOv5 with better performance was selected as the broiler behavior recognition model.This model can recognize broiler broilers with adhesion and occlusion in a cage environment.The accuracy of this model in detecting four behaviors of drinking water,feeding,lying,and standing on the visible light image test set of broilers is 92.5%,94.5%,94.6%,and 91.4%,respectively.The average accuracy m AP(Io U=0.5)is 91.7%.A health monitoring method for broiler flocks is proposed based on the behavior recognition model of broilers.The behavior information and central coordinates of individual broilers in visible light images are obtained,and the changes and position distribution of four behaviors of broilers,namely standing,lying,feeding,and drinking,over time are statistically analyzed.The results show that the lying behavior of healthy broiler flocks significantly decreases within 5 minutes after feeding,while the feeding behavior increases significantly.A study was conducted on the behavioral changes of diseased broiler flocks before and after feeding,and the results showed that there were significant differences in lying and feeding behaviors between diseased broiler flocks and healthy broiler flocks within 5 minutes after feeding.As the number of diseased broilers in the cage increases,the feeding time shows a decreasing trend,while the lying time shows an increasing trend.Comparing the average feeding time and lying time of healthy broiler flocks after feeding with diseased broiler flocks,it was found that the latter had significantly lower feeding time and significantly increased lying time.Therefore,abnormal feeding and lying behavior after feeding can be used as criteria for evaluating the health of broiler flocks,which is of great significance for timely detection of health abnormalities in broiler flocks.(3)Research on health monitoring methods for broiler flocks based on thermal imaging.This study achieved the recognition of dead individuals in broiler flocks based on thermal infrared imaging technology.To improve the accuracy and timeliness of dead broiler recognition,the changes in thermal infrared images of broiler broilers with different postures and parts with time of death were studied.Based on the thermal infrared characteristics of broiler broilers,the judgment basis for dead broiler recognition in thermal infrared images was determined,thereby improving the performance of the dead broiler recognition model.The average accuracy mean m AP(Io U=0.5)of this model on the thermal infrared image test set can reach 95.3%,and the detection speed can reach 69.4fps,meeting the requirements of detection accuracy and speed.(4)Design and testing of a dead broiler removal device.This study designed a dead broiler removal device applied to stacked cages.In order to select and design the hardware part of the device and avoid problems such as damage and fluid outflow caused by compression of dead broilers during the operation of the device,experiments were conducted on the body weight and size characteristics,friction coefficient on the body surface,and compression characteristics of broiler broilers.This device includes a control system and hardware,which removes dead broilers from the cage to the manure cleaning belt through a scraper,and cleans them out of the broiler coop with the operation of the manure cleaning belt.A dead broiler removal experiment was conducted on 4-7 week old broilers,with an average success rate of 83%.