Research On Vision-based Automatic Monitoring Method Of Individual Cow Feeding Information

Precision feeding is an important part of intelligent animal husbandry,and accurate monitoring of individual cow feeding information is a necessary way to achieve the goal of accurate feeding.This study investigates the problems of high cost,low accuracy in traditional dairy cattle feeding information monitoring methods,and proposes a machine vision-based feeding information monitoring method to provide a basis for precision feeding.Based on domestic and international research results,this study aims to improve the accuracy and automation of individual cow feeding information monitoring,and investigates the detection and identification of cow head targets from a feeding perspective,individual cow feeding amount monitoring in a feeding trough scenario,individual cow feeding amount monitoring in a feeding channel scenario,individual cow feeding information identification and quantification,and automatic monitoring of individual cow feeding information based on machine vision.The design and implementation of a system for automatic monitoring of individual cow feeding information based on machine vision is investigated.The main work of this paper is as follows:(1)Research on dairy cow target detection and identification methods.A data acquisition system based on the depth camera Gemini was designed and used to collect RGB images and depth images of cows’ heads from a feeding perspective,to obtain RGB-D data through data fusion and to build an identity dataset.Research the dairy cow target detection method based on YOLO v5 to realize the cow head target detection from the perspective of feeding;in order to solve the retraining problem caused by the increase and decrease of cows in the actual cattle farm,a cow identification method based on deep feature extraction and matching is proposed,firstly,a cow identification feature extractor is constructed based on the residual network;secondly,the cow head features are extracted using the feature extractor,and a feature template library is constructed using the extracted features;finally,the feature matching method is specified to complete the cow identification.(2)Research on individual cow feed intake monitoring methods in the feeding trough scenario.Firstly,a feed pile image acquisition system was constructed for the feeding trough scenario,based on which a dataset was built;secondly,an analysis of cow feeding calculation methods was conducted,and a Siamese network-based individual cow feeding monitoring method was proposed;finally,a multi-layer perceptron is used to quantify the feed intake values to complete the calculation of individual cow feed intake in the feed trough scenario.(3)Research on individual intake monitoring methods for dairy cows in a foraging channel scenario.The method is based on Siamese networks and point cloud data,and is based on the characteristics of the feeding channel scenario in real production.Firstly,a three-dimensional point cloud is generated using the collected feed pile depth data,and automatic clutter removal is carried out by point cloud filtering;secondly,a point cloud data feature extraction model with a selfattentive mechanism is constructed;finally,an individual cow intake monitoring model based on Siamese networks and point cloud data is constructed to complete the individual cow intake calculation in the foraging channel scenario.(4)Design and implementation of a machine vision-based system for automatic monitoring of individual cow feeding information.Firstly,the general design of the system is presented,and the logic of the system and the way of calculating feeding information are given.Finally,a mobile application was developed for data visualization to complete an automatic monitoring system for individual cow feeding systems based on machine vision.Based on computer vision technology,combined with deep learning algorithm,this paper realizes the target detection and identification of dairy cow head from the perspective of feeding.For two different feeding scenarios,a monitoring model of individual dairy cow feed intake is constructed,and a computer vision-based monitoring model is designed and implemented.The visual automatic monitoring system of dairy cow individual feeding information provides technical support for solving key problems in the field of dairy cow refined feeding,and at the same time provides a new idea for monitoring individual dairy cow feeding information through machine vision.