摘要
传统的2维掌纹识别在图像采集时容易受到干湿度、残影和压力等影响,使得其鲁棒性和准确性降低。为解决这些问题,3维掌纹识别技术应运而生。现有的3维掌纹身份认证技术需要将掌纹的特征提取与匹配识别分开进行,不仅延缓了识别时间,更增加了不同方法优化组合的难度。该文提出一种基于曲面类型(ST)与深度学习融合的3维掌纹识别方法。该方法利用ST图像表示3维掌纹特征,并将其作为卷积神经网络(CNN)的输入,实现网络的训练。测试图像可自行提取掌纹图像特征信息并在网络中直接完成识别。实验结果表明,该文方法在公开数据集上得到了99.43%的准确率和28 ms的识别时间,与传统3维掌纹识别方法相比均有提高,实现了3维掌纹的快速高精度识别。
Traditional Two-Dimensional(2D)palmprint recognition is susceptible to the effects of dry humidity,residual image and pressure during image acquisition,which reduces its robustness and accuracy.To solve these problems,Three-Dimensional(3D)palmprint recognition technology is widely studied.The existing 3D palmprint identity authentication technology needs to separate palmprint feature extraction and matching recognition,which not only delays the recognition time,but also increases the difficulty of optimizing the combination of different methods.A 3D palmprint recognition method is proposed based on the fusion of Surface Type(ST)and deep learning.ST images is used to represent 3D palmprint features and to be as input of Convolutional Neural Network(CNN)to realize training.The test image can be automatically extracted the feature information of the palmprint image and complete the identification directly.The experimental results show that the proposed method has an accuracy of 99.43%and a recognition time of 28 ms on the public data set,which has high performance of accuracy and speed compared with the traditional 3D palmprint recognition methods.
作者
张宗华
王晟贤
高楠
孟召宗
ZHANG Zonghua;WANG Shengxian;GAO Nan;MENG Zhaozong(School of Mechanical Engineering,Hebei University of Technology,Tianjin 300130,China)
出处
《电子与信息学报》
EI
CSCD
北大核心
2022年第4期1469-1475,共7页
Journal of Electronics & Information Technology
基金
国家自然科学基金(52075147,51675160)
重大科学仪器设备开发重点专项(2017YFF0106404)。
关键词
3维掌纹识别
曲面类型
深度学习
卷积神经网络
Three-Dimensional(3D)palmprint recognition
Surface Type(ST)
Deep learning
Convolutional Neural Networks(CNN)
