CNN与SVM和ResNet相结合的牛脸识别系统模型研究与实现

Research and realization on cattle face recognition system model based on CNN combined with SVM and ResNet

针对传统的牛个体身份识别需借助外部工具而产生的管理上的安全隐患及对牛身体的刺激和物理伤害问题,提出基于图像方法的牛脸生物特征提取的方法进行身份识别。同时,由于CNN对硬件设施有高要求并且识别精度低、速度慢,以及传统机器学习方法对先验知识的依赖性,考虑到牛脸数据采集困难的小样本问题,以及自然野外光照、视角、距离、运动等客观环境因素与牛脸特征复杂性,提出以CNN为主体且引入ResNet和SVM相结合的牛脸识别与检测的算法与模型,其较传统的CNN网络结构具有训练收敛速度快、识别率高及泛化性强等特点。在Andriod平台下运用该轻量型模型设计了安卓手机APP,完成了方便快捷的实时识别目标,并通过了实验验证,实验结果准确率达95.1%以上。

In order to solve the problems of management security risks,stimulation and physical harm to the body of cattle caused by the traditional identification of individual cattle by external tools,a biometric extraction method of cattle face based on image method is proposed for identification.At the same time,CNN has high requirements for hardware facilities,low recognition accuracy and slow speed,as well as the dependence of traditional machine learning methods on prior knowledge.Considering the small sample problem caused by the difficulty in collecting bovine face data,as well as the complexity of bovine face features due to the objective environmental factors such as natural field light,angle of view,distance and movement,the algorithm and model of bovine face recognition and detection based on CNN and the combination of ResNet and SVM are proposed.Compared with the traditional CNN network architecture,it has the superiority of convergence,recognition rate and generalization.And in the Andriod platform using the lightweight model design and implementation of Android mobile APP,to complete the convenient and fast real-time target recognition,through the experimental verification,the experimental results are analyzed,the accuracy of 95.1%or more.