Research Of Dairy Cattle's Posture Recognition Based On Multi-Sensor Data Fusion

In order to strengthen the informationization,modernization and intelligent construction of China's dairy cattle breeding industry,modern scientific and technological are used to liberate the artificial labor force and improve the accuracy of judging the health status of dairy cows,thereby improving the production efficiency of dairy farming in China.However,in the current research,the scholars only used the acceleration sensor in the research of the cow's posture recognition,which resulted in a single acquisition information.There is also a certain room for improvement in the accuracy of the attitude angle estimation.Aiming at the current research situation,this thesis proposes dairy cattle's posture recognition method based on multi-sensor data fusion.Firstly,in order to solve the single problem of collecting information,a set of data acquisition nodes with multi-sensors is designed and manufactured.Then,in order to improve the accuracy of the existing cow attitude angle settlement method,research of solving method based on PI regulator and complementary filtering is designed.Finally,by designing the cow exercise experiment,the eigenvalues are extracted by the method based on the rough classification for the seven kinds of actions such as feeding,walking and tickling.Through the decision tree method,the cow's posture is identified and classified,and finally the seven poses of the cows are recognized.The main work of the thesis is as follows:1.Design and manufacture a sensor terminal worn on the neck of cows.The sensor terminal is mainly composed of MPU9250 MEMS sensor(including one of three-axis accelerometer,three-axis magnetometer and three-axis gyroscope),STM32F103RBT6 microprocessor,XC6206 power supply,SD card storage module,to complete the collection and storage of cow's sports signals,and then the wireless transmission of the cow's posture data is realized through the Lora gateway.2.In order to solve the problem of low accuracy and poor stability in the current cow attitude angle calculation method,this thesis proposes a new attitude angle fusion algorithm.Firstly,the initial value of the attitude angle is calculated by the acceleration sensor and the magnetometer sensor,and the data fusion is performed by the method of PI adjustment and complementary filtering.Then,the attitude angle is solved by using the quaternion differential equation.Finally,the cow's movement posture is obtained,and the accuracy of the cow's attitude angle calculation is improved.3.Design the cow's exercise experiment.Firstly,the MEMS sensors are used to collect the primitive sports data of the cow,and the collected cow's exercise data is subjected to posture angle solving processing;then the cow's movement posture is classified to determine the feeding,walking,tickling,static,etc,seven kinds of actions are identified.Then,the characteristic parameters are extracted by the rough classification,and the decision tree-based posture recognition method is designed to realize the accurate recognition of the cow's daily movement posture.