Nowadays,more and more people are inclined to smart farming.Sheep face image acquisition has the advantages of non-contact and non-stress to the sheep.Therefore,it is of great practical significance to realize sheep identity management through sheep face image recognition,and the collection of sheep identity information is an important prerequisite to realize sheep individual identity recognition.This paper discusses how to quickly and efficiently collect high-quality sheep face images,analyzes the network structure of YOLOv5 target detection algorithm and the target prediction process,improves the YOLOv5 network from the perspective of multi-scale feature fusion,and constructs a new target detection algorithm DFD-YOLOv5.The detection speed is increased by about 12%.Then,the Alexnet network model was stacked to convert a multi-classification problem into multiple binary classification problems so as to filter the clarity of the image,and the fuzzy sheep face pictures were removed.
data acquisitionsheep face recognitionobject detectionanimal husbandryYOLOv5
国家科技期刊平台
NSTL
维普
万方数据
