Research On Cow Identification Method Based On Body Feature Image And Deep Learning

Identification technology of cows is an important part of automated breeding,which is extremely effective in protecting cows' health and reducing epidemic situation.At present,the most commonly used identification technology of cows is to wear a electronic tags with labels in ear.This method will cause physical damage to cows,and the labels are easily to be lost.Therefore,this paper proposes a identification method of cows based on body feature image and deep learning.The main work of this paper is as follows:(1)build a dairy image acquisition platform,including placing a webcam in the milking parlor and transmitting the video and configuring the operating environment.Produce a cow image dataset that includes extracting cow video frames,sample set preprocessing,manually annotating images,and performing data set enhancements in different ways.(2)the overall framework and function of the cow identification system are analyzed,and the overall detection process of the system is designed.The YOLOv3 detection algorithm is selected to train and test the self-made cow dataset.At the same time,the video of the cow is tested,and the experimental results are analyzed and summarized.Other YOLO algorithms are used to test the same cow dataset and compare the experimental results and detection performance of different YOLO algorithms.(3)in order to improve the detection accuracy of the system,this thesis supplements the two aspects of data sample enhancement and YOLOv3 network structure improvement,and proposes a multi-scale target detection algorithm that joins the spatial pyramid pooling layer.The test results show that the optimized model is more suitable for the identification of cows in this paper.In addition,based on WAMP,a system display platform was designed to realize a complete cow identification system.