Research On The Perception And Classification Of Dairy Cow Ruminating Behavior Based On Multi-source Information Fusion

Dairy farming is the core businesses of modern national agriculture.China's dairy farming,has become an efficient and independent,and made a rapid progress since 1980 s.Rumination is a critical physical digestion process of dairy cows that reflects the health status and estrus,and it has been paid more attention by breeding production.However,due to the limitations of equipment and breeding scale,the development of intelligent monitoring technology for rumination of dairy cows in China is lagged behind,especially the research on intelligent classification methods of rumination data.The dairy cows' rumination behavior monitoring methods include voice,video recorder and activity monitoring.Among these single monitoring methods,there are some shortages(e.g.difficult to analyze and calculate the sound and image information,and hard to distinguish the rumination and feeding behavior only through the behavior monitoring).To achieve an accurate automatic monitoring of dairy cows' rumination behavior,the Chinese Holstein cows were selected as the research animal in this project.According to the features of rumination sound and activity,a novelty multi-source parameter portable neck-wearing monitoring device was developed to achieve a hardware support,and the automatic algorithms of K-means and Support Vector Machine(SVM)were used to realize the intelligent classification of rumination behavior.The device was worn at the best rumination signal monitoring point on the side of the cow neck to monitor the signals of rumination sounds and activities.In order to verify the recognition effect of the wearable device,the comparison was made between monitoring device and manual observation by selecting the 16 representative time nodes of two dairy cows.It is shown that the results from the rumination monitoring device were highly consistent with the manual observations,and the accuracy rate of identifying rumination behavior of the device can reach 81.3%.In order to verify the effectiveness of classification algorithm,the recognition results of SVM and k-means + SVM were compared,the classification accuracy of SVM was between 73.6% and 90.1%,and the classification accuracy of k-means + SVM was between 83.3% and 93.2%.and the axis values of 3D acceleration sensor for regurgitation were 0.09 to 0.93 g for x-axis,-0.16 to-0.08 g for y-axis.The results showed that the wearable device and classification methods proposed in this research were effective for rumination monitoring,and it could provide a feasible reference for the management of dairy farming in the complex breeding environment.