摘要
为了提高指纹识别算法的准确率,文中提出一种基于改进的残差网络的指纹识别算法。该算法首先对指纹图像进行预处理,包括裁剪、旋转等操作,旨在增强数据的多样性;然后设计深度卷积网络,此网络是由多个改进的残差单元连接组成,专为指纹识别而设计,用于提取指纹图像的特征;最后针对深度卷积网络,采用交叉熵作为损失函数,利用Adam算法进行优化。实验结果表明:文中算法的识别准确率达到98.79%,识别时间约为60 ms,模型大小约4.29 MB;与AlexNet及VGG相比,该算法准确率更高,模型更小,在减少处理时间的同时不会过度拟合,可显著提高指纹识别的性能。
A fingerprint recognition algorithm based on improved residual network is proposed to improve the accuracy of fingerprint recognition algorithm. In the algorithm,the fingerprint image is preprocessed including cropping,rotation,etc.,to enhance the data diversity. The deep convolutional network is designed,which is composed of multiple improved residual units connected and designed for fingerprint recognition,so as to extract the features of fingerprint images. Taking the cross entropy as the loss function,the deep convolutional network is optimized by means of the Adam algorithm. The experimental results show that the recognition accuracy of the algorithm is 98.79%,the recognition time is about 60 ms,and the size of the model is about4.29 MB. In comparison with AlexNet and VGG,this algorithm has higher accuracy rate,smaller model,and can reduce processing time but not overfitting,which can significantly improve the performance of fingerprint recognition.
作者
刘晓薇
赵庆东
吴小林
LIU Xiaowei;ZHAO Qingdong;WU Xiaolin(School of Mathematics and Computer Science,Jiangxi Science and Technology Normal University,Nanchang 330038,China;Shenzhen Dotu Technology Co.,Ltd.,Shenzhen 518000,China)
出处
《现代电子技术》
2022年第12期173-176,共4页
Modern Electronics Technique
基金
吉林省教育厅十三五科研规划项目(JJKH20181041KJ)。
关键词
指纹识别
深度学习
卷积网络
图像处理
人工智能
残差网络
交叉熵
fingerprint recognition
deep learning
convolutional network
image processing
artificial intelligence
residual network
cross entropy