Study Of Pig Behavior Monitoring System Based On WSN

With the industrial scale of development of pig production, the most important animal husbandry industry in China, the requirement of husbandry auto-management is put forward to higher level. Due to the limited communication distance and poor extension ability, the commonly-used RFID ear tag cannot meet the requirement of intelligent precise auto-husbandry. Thus, it will advantageous to better the management and increase the income of pig production by incorporating the environment of pig house, the behavior and diet of pigs into monitoring system to track the process timely and dynamically. The main research results are as follows:This dissertation studies several key technologies in behavior monitoring system of pig using the wireless sensor network (WSN). It includs the WSN nodes deployment issue in pig house, the location issue of scatter-feed pigs, the movinig modes monitoring of scatter-feed pigs, and the E-R mode and structure of comprehensive monitoring system for pig production. The descriptions are as follows:1) Peripheral deployment issue in pig house for scatter-feed pigs is studied. The evaluation index system for wireless sensor nodes deployment in pig house is established Evaluation indexes of K-Coverage Ratio, Efficiency of K-Coverage and Minimum radius for K-coverage are put forward based on existing indexes of coverage, efficiency and coverage of variance. An evaluation index system for different nodes deployments is built. Then the system is used to evaluate and optimize four deployed strategies. According to the results, a series of analysis methods of evaluating are created.2) The calculation method of the issue of least number of nodes used for K=1peripheral covering is formed. Two softwares, one for calculating the least number of nodes, the other for simulating the deployment, are designed and programmed. Three deployment examples show the evaluation and the calculation are efficient. Based on Adobe Dreamweaver software development platforms and PHPnow on Internet application framework, a pigsty boundary condition and minimum node calculation software is developed, which can verity the number of the deployment nodes; based on the Matlab software platform, a pigsty node deployment simulation software is developed, which can analyze and simulate the deployment strategies3) A novel piggery trilateral rectangular weighted centroid algorithm is stablished. The measured experiments is studied, using commercial sensor nodes made by different manufacturer. The optimal mode of the function of RSSI and distance within0-80m and0-20m is derived by curve-fitting. On the basis of the trilateral weighted centroid algorithm, a novel piggery trilateral rectangular weighted centroid algorithm is developed, and the procedure is designed further. Experiments of indoor space test in0.43m*5.72m show that the even location error is1.346m, and further shrink to1.0875m when two location data relative with unnormal location are removed from the data set, The time interval for locating is10s.4) The pig moving modes monitoring is realized by three-axis accelerometer A three-axis accelerometer circuit on wireless sensor node is designed. In the wireless sensor node, an software based on Zigbee protocol stack is developed for acquiring acceleration data and sending them to the PC. The protocol for communication is studied.5) A method of moving mode recognition neural network is studied. The two-layer moving mode recognition neural network classifier is built based on the classical theory of the neural network pattern recognition algorithm. A software based on Matlab platform is designed to achieve distinguishing four actions (walk, run, jump, stand up). Simulation experiments proved that the recognition rate for85training samples reachs100%; for439group samples, the recognition rate of neural network method in two moving status, walking and standing up, is still100%, but in running and jumping, recognition rates of the two acts, because the data characteristics more akin to each other, up to99.1%and96.1%respectively, and98.9%of overall resolution is obtained.6) Comprehensive monitoring of swine behaviors are studied. For the first time the algorithm of short-time energy and short-time zero crossing ratio are used to classify the tri-axial acceleration data. Based on VS2010platform, a comprehensive real-time pig moving modes monitoring software is developed. Algorithms used in the software include the short-time energy and short-time zero crossing ratio classification algorithm, fast Fourier translating algorithm and a standard deviation algorithm. Real-time simulated tests proved that the classification rate of the short-time energy classification algorithm is higher than that of FFT fast Fourier classification algorithm and the standard deviation algorithm. The three categories of acting modes, including walking, standing, jumping and running, can be distinguished by the software almost complemently. The recognition rate of jumping mode and running mode is about80%.7) Analysis and design of the pig comprehensive precise monitoring system based on WSN frame structure are performed. Subsequently, E-R model for the networked comprehensive monitoring information system.In all the results in this paper, the hardware design is beased on the method used in automatic control system for grading pipeline of dyeing bamboo sticks; the design of E-R model for the networked comprehensive monitoring information system is built on the basis of Research of network information management system for high-throughput rice breeding. The results provide a scientific solution for application of wireless sensor network (WSN) in pig comprehensive monitoring system. They can lay good foundation for realizing the pig comprehensive accurate monitoring system in future further research works. And they will bring a broad application prospect and development of intelligent auto-husbandry pigs in the future.