Study On The Key Technologies In WMSN Based Average Weight Gain Mornitoring System For Piglets

China is an agricultural country, after 30 years of reforming and opening up, China has ranked first in pig raising quantity in the world. Piglets’healthy breeding is the foundation of pig production. In the period of lactation, the piglets’survival rate and weight gain is what the farmers most concerned about. The weight gain rate of piglets in lactation period is an important index to measure the performance of sows and lactation performance of sows, and an important basis for farmers to select quality sow. The weight gain rate of piglets in lactation period also affects their growth in conservation period and fattening period directly. Piglets with poor weight gain in lactation period have more difficulty during weaning, have higher mortality in nursery stage, slower weight gain in fattening period and lower feeding efficiency. At present, people generally weigh piglets twice, at birth and weaning, which requires a lot of human resources and puts great stress on piglets. Frequent movements of piglets on the scale also affect the weighing accuracy. Therefore, farms need a system, which can real-time, continuously and automatically monitor lactation piglets weight information. What farm staff can implement based on the information is an artificial intervention for newborn pigs with abnormally weight gaining to promote healthy growth of piglets in lactation, and to optimize economic benefits of the farms by selecting quality gilts.In view of the actual demand of automatic, continuous and non-contact monitoring system by scale raise fields on piglet, this thesis designs an automatic monitoring system of piglets’nest average weight gain in lactation period with wireless multimedia sensor network, and solves the key technical problems in system design. This issue is studied in the following aspects:(1)Put forward the framework of piglets’ average weight gain monitoring system based on wireless multimedia sensor network, designed the image sensor node which was equipped with the load cell, CMOS image sensor and wireless RF module. Improved the current piglets’incubators in the following aspects:increase the LED light source which can provide a stable light condition for image sensor nodes collection box piglets image illumination. Replaced the traditional heat lamps by the insulation board, so that the temperature distribution in the incubators will be more uniform, and then the probability of the piglets serious overlapping resting posture is reduced effectively, which will be benefit for the piglets counting in an image by image segmentation.(2)In order to monitor the piglets’average weight automatically in lactation period, designed a low power wireless multimedia sensor network node, introduced the sleep wake-up mechanism, which greatly reduced the energy consumption of nodes. Nodes with low power STM32 as micro controller, CC1101 as the RF module, extended OV7620 image sensor and the weighing sensor. When weighing sensor under the bottom plate of piglets incubator output stable weighing signal acquisition, OV7620 collects the piglets’ image in incubator. Multimedia sensor nodes package piglets’image according to the frame format of Simplici TI protocol and send to the sink node. The server collects the image through the sink node and determines the number of piglets automatically using image segmentation algorithm, combined with automatic monitoring implementation of piglet average heavy weight data.(3) Image sensor nodes of wireless multimedia sensor networks in agriculture usually adopt single chip MCU and without extended SRAM for the reason of low price. But the image data is too large for the low bandwidth WMSN, the images have to be compressed before transferred through the WMSN. This brings a big challenge to the MCU which is low memory and without the floating point arithmetic unit. This thesis adopted a line based wavelet transform method based on Le Gall 5/3 filter, proposed a bit plane binary adaptive arithmetic coding algorithm for the wavelet efficient based on binary adaptive arithmetic coding method. Combined with the line based multilevel wavelet transform, quantization and coding, designed a low memory, low complexity, and efficient image compression algorithm which is suitable for single chip micro controller.(4) Acquired the temperature compensation coefficients for the load cell based on artificial neural network, resolved the problem of the nonlinear compensation for the load cell. The environmental parameters of breeding buildings are complex:high humidity and high temperature in summer, cold in winter. It doesn’t satisfy the linear relationship between the output voltage of the sensor and the pressure because the range of the temperature in the breeding buildings is large, which in turn affects the accuracy of the weigh results. This thesis studied the temperature compensation method for piezoresistive load cell based on the training sample data and BP neural network, improved the measurement accuracy of the weigh result.(5) In the application of piglets’ average weight monitoring, counting the number of the piglets in an automatic manner is an important research content because the rest poses of piglets in the incubator are many and varied. This thesis proposed an image segmentation algorithm for the adhesive piglets based on the multi-ellipse fitting method. First, execute the ellipse fitting operation for a large number of images which have individual piglet and extract the key parameters of fitting ellipse for piglets of different ages. Then extract the contour of the adhesive piglets’ image, segment the contour based on the concave points which are extracted based on K chain code. Execute the ellipse fitting operation for every contour segment. Finally,5 kinds of elliptic merge principle was put forwarded, which were used to merge the anomalous elliptic. The number of the adhesive piglets in an image is equal to the number of ellipses after the merger operation.Designed prototype system for piglets’ average weight monitoring and test the system in the laboratory, the results of the test show that the prototype system can monitor the average weight of piglets in incubator remotely in a high precision. Finally the thesis summarized the research on the subject, and prospected the work in the future.