Pig Individual Information Acquisition And Behavior Recognition Method Research

Pig breeding occupies a major proportion in Chinese livestock and poultry breeding industry.Large-scaled farming is an inevitable trend in the development of modern pig breeding industry.Pig health real-time monitoring has become a research focus in the modern breeding technology.Test show that pig behavior,temperature and pulse could directly reflect its health condition.Animal recognition model based on existing methods of machine vision,image analysis,RFID,which are complex segmentation algorithm,low real-time performance and high equipment costs,are not suitable for a wide range of application.Pig daily behavior corresponding of attitude angle change law was analyzed in the paper.A wearable pig behavior information retrieval platform was built by adopting low power consumptionSTC12C5A60S2 core processor and multi-sensor information technology.Then,the fusion attitude angle of pig daily behavior recognition model was established in this paper.Multi-source information fusion of pig physical activity and sleeping time statistical model were established by analyzing triaxial acceleration information when the pig was walking,combination with pig body temperature and heart rate information.The PC system wasdesigned based on above models,and realized pig daily behaviors in real-time monitoring.Finally,the Guanzhong wild pigs and Changbai sows were selected to test system function and stability.In a conclusion,the wearable pig information acquisition device,which operated steady and possessed a high accuracy,opens up a new research direction for real-time monitoring large-scale individual pig physical indexes.It's beneficial to speed up the transformation of Chinese pig industry to high-end market.It provides theoretical basis and data support to analyze individual pig physical condition and provides the promising reference for related animal health monitoring.In this paper,the main work and conclusions were as follows:(1)Theoretical foundation of pig behaviorrecognition was analyzed,and the overall function of the monitoring system was designed.Triaxial acceleration and angles as the key factor of the recognition pig different behaviors were identified by analyzing the change law of attitude information when the pig different behaviors occur.Dynamic Kalman filtering and mean filtering algorithm with calibration and optimization of gyroscope was adopted to eliminate gyroscope random drift of attitude angle calculating accumulative error.The result of attitude accuracy reached 0.01°.Considering of piggery environment,the pig daily behaviors monitoring system was designed,which was composed of behaviors information acquisition device and upper machine.The device was tied to the neck of the pig with a loose bag.The above information was transmitted to upper machine by wireless Bluetooth.Finally the statistics and analysis of the individual pig daily behaviors information were realized.(2)Setting up a wearable device of pig movement information acquisition.The device was composed with low-power core processor STC12C5A60S2 and the device could real-time collect pig behaviors,temperature and pulse information.The underlying design mainly includes behaviors information device packing design,the overall hardware circuit design,the multi-sensor acquisition module design,and wireless Bluetooth transmission.The device is wrapped in 3D printing technology.The overall size is 65mm×25mm×30mm and reduces device volumeeffectively.This paper provides the basis for the optimal design of the wearable pig information acquisition.(3)Pig behavior recognition method with attitude angle based on BP neural network was proposed in this paper.Pig daily four behaviors corresponding attitude information were acquired by constructed the pig information collection platform.The data was the input vector of BP neural network,compared different BP network parameters,to determine the optimal network structure.LM training method was used to respectively construct pig behavior recognition model of fusion and non fusion attitude angle.The recognition model with attitude angle error was less than 0.0018 and convergence rapidity.Pig exercise and sleeping time statistical model were built based on multi-source information fusion body by considering temperature and heart rate change rule combining with pig specific behavior.The results showed that the accuracy behavior discriminant rate was 90.47%,the amount of exercise and sleeping time statistical accuracy were 84.28% and 86.61%.(4)Pig behaviors monitoring system was constructed based on C# language.The monitoring system was developed by using C# programming language and visual studio 2010 platform and the behaviors classification and statistical model were embedded into the PC system.Then,the system integral function andkey parameter indexes were measured.It's showed that the system could continue working about 7.6 days and the overall function operated smooth.It provided a guarantee for real-time monitoring pig daily behaviors.