基于改进的InceptionV3模型在肉牛体侧识别中的应用研究  

Research on the application of the improved InceptionV3 model in body side recognition of beef cattle

作　　者：张宇[1] 李宝山[1] 高迪 杜永兴[1] ZHANG Yu;LI Baosan;GAO Di;DU Yongxing(School of Digital and Intelligent Industry,Inner Mongolia University of Science&Technology,Baotou 014010,China)

机构地区：[1]内蒙古科技大学数智产业学院,内蒙古包头014010

出　　处：《内蒙古科技大学学报》2024年第4期365-370,共6页Journal of Inner Mongolia University of Science and Technology

基　　金：国家自然科学基金(61961033);内蒙古自治区自然科学基金(2019MS06021);内蒙古自治区科技重大专项(2019ZD025)。

摘　　要：为了实现无接触,高精度复杂养殖环境下对无明显特征肉牛个体的有效识别,实验基于InceptionV3模型构建了结合迁移学习和改进网络结构的InceptionV3识别模型,利用该网络模型对肉牛体侧数据集进行训练。结果表明:基于改进的InceptionV3模型对养殖场中肉牛的识别准确率为99.70%,具有较高的识别准确率。说明利用深度学习方法构建的算法模型对肉牛个体进行身份识别能满足养殖场的需求,可应用于肉牛个体识别领域。In order to achieve effective identification of individual beef cattle without obvious features in a contactless,high-precision,complex breeding environment,based on the InceptionV3 model,an InceptionV3 recognition model combining transfer learning and improved network structure was constructed,and the network model was used to train the beef cattle body side data set.The results show that the recognition accuracy of the improved InceptionV3 model is 99.70%for beef cattle on the farm,which has a high recognition accuracy.It indicates that the algorithm model constructed by using deep learning methods for the identification of individual beef cattle can meet the needs of farms and can be applied in the field of individual beef cattle identification.

关 键 词：深度学习 肉牛个体识别 迁移学习 改进网络结构 InceptionV3 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]