Research On Real-time Monitoring Of Dairy Cow Rumination Behavior Based On Edge Computing

As China’s dairy farming industry rapidly develops towards large-scale,precision,and information-based farming,the demand for monitoring and management of dairy cow health is increasing.Timely monitoring of rumination behavior in dairy cows is essential for obtaining relevant information on their health and predicting diseases.Currently,various strategies have been proposed for monitoring rumination behavior,including video surveillance,sound recognition,and sensor monitoring.These methods can identify rumination behavior in cows,but they generally lack real-time performance.To address this issue,this study investigated two real-time methods for monitoring rumination behavior in dairy cows in different levels of intelligent farming environments.The specific research content includes the following four aspects:(1)The overall architecture of the system based on edge computing and edge intelligence.For cattle farms with relatively weak infrastructure,a decentralized edge intelligence model is proposed for the three-axis acceleration data collected by three-axis sensors,and a cloud-end integrated system is designed based on this model.By assigning computing and tasks to the edge for processing,not only can effectively reduce data transmission and computing load,but also effectively avoid recognition delays and errors caused by network delays and bandwidth limitations;for large-scale precision livestock farms,in order to achieve The effective expansion of the data dimension of dairy cows enables deeper feature acquisition,thereby improving the recognition accuracy and monitoring efficiency of rumination behavior,and at the same time providing a data basis for the research of monitoring other behaviors.This paper adds three-axis angular velocity data on the basis of threeaxis acceleration data.Build six-axis data.For the six-axis data collected by the six-axis sensor,the edge layer is introduced to study the federated and split edge intelligent systems based on the sixaxis sensor,and a cloud-edge collaborative system is designed.(2)Real-time recognition of cow rumination behavior based on triaxial sensors.For dairy farms with relatively poor network infrastructure and sensitive production costs,this paper uses selfdesigned edge devices to collect and process cow’s triaxial acceleration signals in real-time.Then,using a rumination recognition algorithm based on K-nearest neighbor,the overall sliding geometric mean of the Euclidean distance between the feature data sets is calculated in real-time to determine the adaptive threshold.The rumination behavior is verified using a sliding window,and a decentralized edge intelligence model is constructed.The average precision rate,recall rate,and F1-score of the cow rumination behavior recognition of this model reached 93.7%,92.8%,and 93.4%,respectively.Finally,real-time recognition of cow rumination behavior is achieved at the edge device end,and real-time calculation can be completed without requiring a large amount of computing time and resources.(3)Real-time recognition of cow rumination behavior based on six-axis sensor.For large-scale farms with high requirements for informatization and precision,this paper proposes two different strategies based on federated and split edge intelligence based on six-axis data to conduct research on real-time recognition of cow rumination behavior.In the research on the real-time recognition method of cow rumination behavior based on federated edge intelligence,the CA-Mobile Net network was proposed by improving the Mobile Net v3 network through the collaborative attention mechanism,and then the federated edge intelligence was designed using the CA-Mobile Net network and the Fed Avg model aggregation algorithm Model.In the research on the real-time recognition method of dairy cow rumination behavior based on split edge intelligence,a split edge intelligence model based on Mobile Net-LSTM was designed by using the Mobile Net v3 network and Bi-LSTM network with the cooperative attention mechanism.Through comparative experiments,the average precision,recall,F1-score,specificity and accuracy of CA-Mobile Net-based federated edge intelligence model reached 97.1%,97.9%,97.5%,98.3% and 98.2%,respectively.The best recognition effect is achieved.(4)Real-time monitoring system of cow rumination behavior based on edge computing.Design and develop a cloud-end integrated system based on a three-axis sensor and a decentralized edge intelligence model;design and develop a cloud-edge collaboration system based on a six-axis sensor and a federated edge intelligence model.The cow’s rumination behavior is recognized at the edge,and the rumination information is finally uploaded to the cloud server.By building and verifying the real-time monitoring system of dairy cow rumination behavior based on edge computing,the frequency and duration of dairy cow rumination behavior are summarized and visualized,and the real-time monitoring of dairy cow rumination behavior is realized.In summary,this paper proposes an effective monitoring method for cow rumination behavior based on edge computing technology,and realizes real-time and accurate monitoring of cow rumination behavior for different edge computing environments.This technology provides support for the promotion of individualized,precise,and information-based dairy farming,and also provides a new method for dairy cow health monitoring and early warning management.