Research On Automatic Scoring Technology Of Dairy Cow Body Condition In Edge Computing Environment

In the process of dairy cow breeding management,body condition score(BCS)is an important tool to evaluate the production performance and health status of dairy cows,which is of great significance to the development of precision husbandry.Due to the low efficiency and high subjectivity of manual scoring,body condition scoring method based on artificial feature extraction has poor robustness in complex scenes because the features are manually defined and selected,which is empirical and subjective.The body condition scoring method based on deep learning can extract the body condition related features more efficiently and automatically.The body condition of dairy cows can be scored automatically through the end-to-end body condition scoring model.This paper researched the automatic scoring technology of dairy cow’s body condition in edge computing environment.To achieve automatic and precise estimation of dairy cow body condition,firstly,dairy cow object detection is carried out on the input image,and dairy cow individual identification was carried out based on the detection results,then the dairy cow body condition score was estimated.Finally,an automatic scoring system of dairy cow’s body condition in edge computing environment was developed which was proposed to solve the issues of the experience of the traditional body condition scoring method and the low efficiency of the dairy cow body condition automatic scoring system based on PC.The main research contents of this paper are as follows:(1)Research on object detection and individual identification of dairy cows.Dairy cow object detection and individual identification are the primary work of precision body condition scoring.This paper collected dairy cow RBG image and depth image data through the data acquisition device with the depth camera Intel Realsense D435 as core.In order to enable the model to carry out normal reasoning in edge computing environment,in terms of dairy cow object detection,the dairy cow object detection method based on YOLOV4 was researched.In the aspect of dairy cow individual identification,the dairy cow individual identification method based on Mobile Net V2 was researched.(2)Research on automatic scoring method of dairy cow body condition.This paper researched the automatic scoring method of dairy cow body condition based on two different strategies.In the research on the automatic scoring method of dairy cow body condition based on depth image feature extraction,the depth histogram was used to process the depth image,and the three-channel dairy cow depth image fused with edge detection of RGB image and Fourier transform of depth image was used as input to research the automatic scoring model of dairy cow body condition based on Shuffle Net V2.In the research of dairy cow condition scoring method based on three-dimensional feature extraction,point cloud data was used as research object,and an attention guided point cloud feature extraction network was proposed to estimate the dairy cow body condition.(3)Design of edge computing oriented automatic scoring system for dairy cow body condition.In the overall design of the system,Intel Realsense D435 camera was used as the terminal layer to collect dairy cow image data.The NVIDIA Jetson AGX Xavier device was selected as the hardware of the edge computing layer.After the environment of the device was built,the dairy cow object detection,individual identification and automatic body condition scoring model were integrated and transplanted to the device.Finally,the output individual identification and body condition scoring results were uploaded to the cloud server through HTTP protocol.Then an automatic and accurate dairy cow body condition automatic scoring was achieved at the end close to the data source.Based on deep learning technology,this paper achieved the automatic and precise estimation of dairy cow body condition score in edge computing environment,and completed the design of dairy cow body condition automatic scoring system in edge computing environment,which not only provided technical support for achieving the goal of personalized,accurate and intelligent dairy cow farming management,but also provided guiding ideology for other intelligent farming fields...