Font Size: a A A

Desin And Implementation Of Cattle Face Recognition Algorithm

Posted on:2022-05-20Degree:MasterType:Thesis
Country:ChinaCandidate:F S MengFull Text:PDF
GTID:2493306509956319Subject:Control Engineering
Abstract/Summary:PDF Full Text Request
With the development of animal husbandry,the scale-up,refinement and intelligent breeding of cattle has become the consequential development trend.The techniques for individual identification of cattle are fundamental to the refinement of cattle cultivation and insurance industry.And efficient techniques for individual identification are needed for cattle breeding,prevention and control of diseases,intelligent management,and regulation of the quality of dairy and beef products.With the development of deep learning in the field of image,the technique of individual identification in cattle based on biological features has made great progress.In this paper,based on the facial features of cattle,a method of cattle facial image recognition based on deep learning has been proposed.The specific work is as follows:1.Building the dataset of cattle facial images.Cattle facial image acquisition of Holstein cows with Simmental beef cattle was performed from April to October 2020 in a pasture in Hohhot City,Inner Mongolia.A total of 50 adult beef cattle and 80 cows were collected.The acquired images were filtered,cropped,and a cattle facial recognition image dataset was constructed.2.Building the cattle facial recognition model based on improved ResNeXt network.The SE(Squeeze-and-Excitation)module is added to the ResNeXt module to construct the SE-ResNeXt cattle facial recognition model.Strengthen the feature extraction ability of the network through the channel attention mechanism.In this way,the accuracy of cattle facial recognition is improved.3.The recognition model of cattle face based on two-branch convolutional neural network is established.The two input images are separately feature extracted through two feature extraction channels.After the extracted features are fused,they are input to the classifier for recognition.In order to reduce the influence of cattle posture change on the recognition accuracy,two images were used to identify individual cattle.4.Detection and recognition visualization of cattle facial images.Build a cattle face detection and recognition interface.Automatically detect the cattle face area,and use the constructed cattle face recognition model to recognize it.Finally,the recognition result is displayed on the interface.
Keywords/Search Tags:convolutional neural network, cattle face recognition, ResNeXt, two-branch convolutional neural network
PDF Full Text Request
Related items