Study On The Monitoring System Of Sheep’s Grazing Behaviors

Animal behaviors are the results of the interaction between animals and their environment.Generally,through analyzing various behaviors of an animal,its health status can be observed to guide how to take advantage of its surroundings effectively.As is known to all,sheep farm is an important animal husbandry in the western region of China.The flock of sheep feeding grassland resources will lead to a couple of grazing behaviors,which are closely related to livestock production,pasture resources and grassland ecology.In this dissertation,the research on the monitoring system of sheep’s grazing behaviors was carried out to identify and classify three grazing behaviors of sheep feeding,chewing and rumination.Also,the feeding path of the flock was monitored with GPS.The research can provide a basic evidence for establishing a better grazing system,as well as improving the grazing management.This dissertation takes the Mongolian sheep as the research targets,whose signals of grazing behaviors including feeding,chewing and rumination were detected under the condition of free grazing.Meanwhile,GPS trajectory information was collected.Then,the wavelet threshold algorithm was put forward to eliminate noise from original signals of grazing behaviors,as well as a BP neural network algorithm model was established to recognize behaviors described above.The following conclusions were obtained:(1)In regard to sheep grazing behaviors,a data acquisition method was proposed.To be concrete,taking MPU6050 tri-axial acceleration sensor as the core,a wearable data collection device was designed to realize the data of sheep grazing behaviors to be real-time collection and wireless transmission.Besides,the data can be saved and visually rendered on the host computer.(2)In term of the acceleration signals of grazing behavior,the noise reduction was achieved by using the wavelet threshold denoising method with the wavelet basis function sym5 and the heuristic threshold(HeurSure).Moreover,the reconstructed acceleration signals were demonstrated to be better with the soft threshold plus 2 decomposition layers.(3)By the aid of the software MATLAB2019a,the signals of grazing behaviors were cut into three kinds of specific signals represented by feeding,chewing and rumination.Then,A BP neural network model was designed to recognize and classify the signals of feeding,chewing and rumination.As a result,the average recognition rate is 88.05%,which can meet the requirements of the recognition of sheep grazing behaviors.(4)As for the flock movement,the GPS positioning system is used to collect the spatiotemporal trajectory,which can be analyzed to get information such as the flock movement speed and the feeding area.