Dorsal hand vein recognition based on transmission-type near infrared imaging and deep residual network with attention mechanism
Dorsal hand vein recognition based on transmission-type near infrared imaging and deep residual network with attention mechanism
复制标题
基于透射式近红外成像和注意力机制深度残差网络的手背静脉识别
DOI：
10.1007/s10043-022-00750-3
复制
发表时间：
2022-06
期刊：
影响因子：
1.2
通讯作者：
Chuncheng Zhang
中科院分区：
文献类型：
--
作者：
Zhenghua Shu;Zhihua Xie;Chuncheng Zhang
手背静脉识别利用人体静脉分布结构的生物信息，具有唯一性、保密性强、防伪能力强等优点。采用常规反射式近红外方法进行手背静脉识别的主要问题是分辨率低、类内变异大、样本不足。针对这些问题，提出了一种基于带注意力的深度残差网络(DRNAM)的手背静脉识别系统。具体地说，改进的手背静脉成像是利用透射式近红外光谱设计的。然后，利用DRNAM模型从手背静脉图像中提取紧凑且具有区分性的特征，通过跨通道和空间信息融合提高了特征提取的鲁棒性。最后，该方法通过对模型的迭代训练获得识别结果。简言之，基于近红外光谱成像的DRNAN模型能够有效地识别手背静脉。实验结果表明，基于透射型近红外光谱的手背静脉图像比基于反射型近红外光谱的手背静脉图像更加清晰，提出的基于DRNAM的手背静脉识别方法优于传统的卷积神经网络方法。
Dorsal hand vein recognition, exploiting the biological information of human vein distribution structure, has the superiority of uniqueness, strong confidentiality and strong anti-forgery ability. The main challenges of the efficient dorsal hand vein recognition via usual reflection-type near infrared means are low resolution, big intra-class variation and insufficient samples. To address these issues, this paper proposed a novel dorsal hand vein recognition system based on the deep residual network with attention mechanism (DRNAM). Specifically, the improved dorsal hand vein imaging is designed by the transmission-type near infrared spectrum. Then, the system aims to extract compact and discriminative features from the dorsal hand vein image by the DRNAM model, which improves the robustness of feature extraction via cross channel and spatial information fusion. Finally, the method achieves the recognition result through the iterative training of the model. Briefly, the DRNAN model could effectively recognize the dorsal hand vein based on near infrared spectral imaging. The experimental results demonstrate that the dorsal hand vein image based on transmission-type near infrared spectrum is clearer than that based on reflection-type near infrared spectrum, and the proposed dorsal hand vein recognition method based on DRNAM outperforms the works based on the traditional convolutional neural network.
登录
查看更多内容
DOI：
10.1007/978-0-387-09766-4_2306
发表时间：
2011
期刊：
--
影响因子：
--
作者：
通讯作者：
--
DOI：
10.1109/tpami.2012.89
发表时间：
2013-01-01
影响因子：
23.6
作者：
Borji, Ali;Itti, Laurent
通讯作者：
Itti, Laurent
DOI：
10.1109/bibm.2017.8217830
发表时间：
2017-11
期刊：
2017 IEEE International Conference on Bioinformatics and Biomedicine (BIBM)
影响因子：
--
作者：
Haipeng Wan;Lei Chen;Hong Song;Jian Yang
通讯作者：
Haipeng Wan;Lei Chen;Hong Song;Jian Yang
影响因子：
3.1
作者：
Jun Wang;Guoqing Wang
通讯作者：
Jun Wang;Guoqing Wang
影响因子：
3.9
作者：
Huafeng Qin;M. El-Yacoubi;Jihai Lin;Bo Liu
通讯作者：
Huafeng Qin;M. El-Yacoubi;Jihai Lin;Bo Liu