Preliminary Exploration On Digitization Of Laying-hen Breeding Systems Based On The Internet Of Things And Streaming Computing

According to the requirements of healthy poultry industry,this study was to explore the transformation of poultry farms from traditional labor-extensive management to precision management.The physiological data,enviormental data and production process data were to be processed in real-time and then the standard data provided by manufacturers were to be compared,in order to improve the poultry farms' capacity of early warning and decision-making.To accomplish the above objectives,some key technologies have been developed,including real-time acquisition and transmission of laying-hen breeding data based on the Internet of Things,real-time processing and analysis technology of laying-hen breeding data based on the distributed streaming computing,and automatic tracking and behavior recognition algorithm for laying hens based on the machine learning technique.These technologies have been integrated into the digital intelligent monitoring and remote management system for laying hens which has been used in chicken farms.The system was based on a distributed architecture and supports management of remote farms.Implementation of the system was to significantly advance the ability of modern poultry producers in their efficient management of large data set,record-keeping,and real-time visualization of the economic performance and housing conditions.Moreover,automatic analysis of video data has been fulfilled and the behavior of laying hens was also visualized.The application of the system in poultry farms has achieved a suitable perception of the breeding situation,which will greatly contribute to the improvement of automation and scientific management,and reduce the intensive labor requirement of China's poultry industry.It's also helpful in well-being farming and real-time decision making of China's poultry industry.The main conclusions are as follows:(1)Through the literature search and investigation on the requirements of multiple chicken farms,three categories of data which impact health and production of laying hens have been summarized,including enviormental data,physiological data and production process data.Aiming at the different characteristics of the data,Internet of Things model was used,which made the perceived data acquired automatically,and other paper-based reports were digitized and stored in the system.An app was developed and provided to enable the staff to operate the system anywhere on the farm with a mobile phone.As the internet infrastructure of chicken farms is still not fully developed in some remote regions,a mechanism of asynchronous transfer was proposed to save bandwidth and data traffic,and to avoid data loss and retransmission because of this network.(2)To implement real-time process,analysis and early warning,"Data-Canal" framework based on distributed streaming computing was designed and developed.Data-Canal is a data flow oriented distributed computing framework with the control flow concentration and data flow dispersion model which uses the distributed file system as the storage of intermediate results to enhance the reliability and scalability greatly.The result showed that the system solves the problem of informatization and real-time processing of mass data in laying-hen breeding systems.In the case of eight machines,the highest throughput of Data-Canal cluster reaches 160 MB/s,and the delay is at the minute level.(3)Video data have become the main part of mass data in laying-hen breeding systems.Due to the lack of recognition,it's difficult to get valuable information from the large number of video segments.Automatic tracking and behavior recognition algorithm for laying hens based on machine learning was developed to analyze the behavioral video data of laying hens automatically.HOG is used to extract the features of the sample laying-hen,and SVM model is used to search the best position of the laying-hen.The algorithm can track any laying hen in a flock,and calculate its movement distance and speed,and quantify its movement behavior.Combined with regional distribution of the facilities,the usage of different areas can be automatically calculated to quantify the behaviors of food intake and water drinking indirectly.(4)This study focused on large-scale laying-hen breeding systems.Under the Cloud architecture,the Internet of Things and distributed streaming computing were used.Automatic data acquisition,asynchronous transmission and processing of real-time data in laying-hen breeding production process were implemented.Combined with the automatic processing of video data based on the algorithm of tracking hens,all data analysis services have been applied in actual poultry farms,i.e.,the economic benefit analysis,real-time warning of production process,environmental data analysis,multidimensional data analysis,production information management,and so on.