双重注意力机制下的跨光谱虹膜识别优化算法

doi:10.3778/j.issn.1002-8331.2106-0116

摘要/Abstract

摘要： 跨光谱虹膜识别任务，通常是指匹配不同光谱下采集的虹膜图像，针对光谱域变化对虹膜识别率影响的问题，提出一种基于双重注意力机制下的跨光谱虹膜识别优化算法。该算法首先通过浅层网络的特征提取，将可见光与近红外光谱域中的浅层特征送入各自的专有外部注意力模块进行优化，以分别存储各自光谱域中虹膜样本数据集的特有信息，随后将优化后的浅层特征送入共享的深度特征提取网络，再将深度特征先后送入改进的空间注意力模块和外部注意力模块，强化虹膜关键特征信息的表达，同时以改进的加性角间距损失函数ArcFace Loss（additive angular margin loss）动态调整对于困难样本的优化力度。该算法在PolyU数据集上进行了实验验证，等错误率EER（equal error rate）达到了0.23%，分离度达到了8.56，该实验结果与跨光谱虹膜识别SOTA（state-of-the-art）算法进行了比较，远高于其他主流算法。

关键词: 跨光谱, 虹膜识别, 注意力机制, ArcFace Loss, 困难样本

Abstract: The task of cross spectral iris recognition usually refers to matching iris images collected under different spectra. Aiming at the problem of the influence of spectral domain change on iris recognition rate, an optimization algorithm of cross spectral iris recognition based on dual attention mechanism is proposed. Firstly, through the feature extraction of shallow network, the shallow features in the visible and near-infrared spectral domain are sent to their own dedicated external attention module for optimization, so as to store the unique information of iris sample datasets in their respective spectral domain, and then the optimized shallow features are sent to the shared deep feature extraction network, Then, the deep features are sent to the improved spatial attention module and external attention module to enhance the expression of iris key feature information. At the same time, the improved additive angular margin loss function（ArcFace Loss） is used to dynamically adjust the optimization strength for difficult samples. Experimental results on PolyU dataset show that EER（equal error rate） is 0.23% and resolution is 8.56. The experimental results are compared with SOTA（state-of-the-art） algorithm, which is much higher than other mainstream algorithms.

Key words: cross spectrum, iris recognition, attention mechanism, ArcFace loss, difficult sample

任家润, 沈文忠. 双重注意力机制下的跨光谱虹膜识别优化算法[J]. 计算机工程与应用, 2023, 59(1): 187-198.

REN Jiarun, SHEN Wenzhong. Optimization Algorithm of Cross Spectral Iris Recognition Based on Dual Attention Mechanism[J]. Computer Engineering and Applications, 2023, 59(1): 187-198.

参考文献

[1] 李海青，孙哲南，谭铁牛，等.虹膜识别技术进展与趋势[J].信息安全研究，2016，2（1）：40-43.
LI H Q，SUN Z N，TAN T N，et al.Progress and trends in iris recognition[J].Information Security Research，2016，2（1）：40-43.
[2] 周锐烨，沈文忠.PI-Unet：异质虹膜精确分割神经网络模型的研究[J].计算机工程与应用，2021，57（15）：223-229.
ZHOU R Y，SHEN W Z.PI-Unet：Research on precise iris segmentation neural network model for heterogeneous iris[J].Computer Engineering and Applications，2021，57（15）：223-229.
[3] 刘京.异质虹膜图像的鲁棒识别[D].合肥：中国科学技术大学，2014.
LIU J.Robust recognition of heterogeneous iris images[D].Hefei：University of Science and Technology of China，2014.
[4] 彭超，徐波.VR眼镜的感观控制方法及VR：CN107516033A[P].2017-12-26.
PENG C，XU B.Perception control method of VR glasses and VR：CN107516033A[P].2017-12-26.
[5] The Hong Kong Polytechnic University cross-spectral iris images database[DB/OL].[2021-05-15].http：//www4.comp.polyu.edu.hk/~csajaykr/polyuiris.htm.
[6] ZUO J，NICOLO F，SCHMID N A.Cross spectral iris matching based on predictive image mapping[C]//Proceedings of the 4th IEEE International Conference on Biometrics：Theory，Applications and Systems（BTAS），2010：1-5.
[7] OKTIANA M，MUNADI K，ARNIA F，et al.Features for cross spectral image matching：A survey[J].Bulletin of Electrical Engineering and Informatics，2018，7（4）.
[8] OKTIANA M，SADDAMI K，ARNIA F，et al.Advancesin cross-spectral iris recognition using integrated gradientface-based normalization[J].IEEE Access，2019，7：130484-130494.
[9] NALLA P R，KUMAR A.Toward more accurate iris recognition using cross-spectral matching[J].IEEE Transactions on Image Processing，2017，26（1）：208-221.
[10] WANG K，KUMAR A.Cross-spectral iris recognition using CNN and supervised discrete hashing[J].Pattern Recognition，2019，86：85-98.
[11] WANG K，KUMAR A.Toward more accurate iris recognition using dilated residual features[J].IEEE Transactions on Information Forensics and Security，2019，14（12）：3233-3245.
[12] MOSTOFA M，TAHERKHANI F，Dawson J，et al.Cross-spectral iris matching using conditional coupled GAN[J].arXiv：2010.11689，2020.
[13] ZANLORENSI L A，LUCIO D R，BRITTO A，et al.Deep representations for cross-spectral ocular biometrics[J].IET Biometrics，2019，9（2）：68-77.
[14] WOO S，PARK J，LEE J Y，et al.CBAM：Convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision（ECCV），2018：3-19.
[15] GUO M H，LIU Z N，MU T J，et al.Beyond self-attention：External attention using two linear layers for visual tasks[J].arXiv：2105.02358，2021.
[16] DENG J K，GUO J，XUE N N，et al.ArcFace：Additive angular margin loss for deep face recognition[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition（CVPR），2019：4685-4694.
[17] WANG H，WANG Y，ZHOU Z，et al.CosFace：Large margin cosine loss for deep face recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2018：5265-5274.
[18] ZAGORUYKO S，KOMODAKIS N.Learning to compare image patches via convolutional neural networks[C]//Proceedings of 2015 IEE Conference on Computer Vision and Pattern Recognition，2015：4353-4361.
[19] HE K，ZHANG X，REN S，et al.Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2016：770-778.
[20] YE M，SHEN J，LIN G，et al.Deep learning for person re-identification：A survey and outlook[J].arXiv：2001. 04193，2020.
[21] FILIP R，GIORGOS T，ONDREJ C.Fine-tuning CNN image retrieval with no human annotation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2017，41：1655-1668.
[22] ZHANG H，GOODFELLOW I，METAXAS D，et al.Self-attention generative adversarial networks[J].arXiv：1805. 08318，2018.
[23] HERMANS A，BEYER L，LEIBE B.In defense of the triplet loss for person re-identification[J].arXiv：1703. 07737，2017.
[24] 袁亚湘，孙文瑜.最优化理论与方法[M]北京：科学出版社，2001.
YUAN Y X，SUN W Y.Optimization theory and methods[M].Beijing：Science Press，2001.
[25] WILD P，RADU P，FERRYMAN J.On fusion for multispectral iris recognition[C]//Proceedings of International Conference on Biometrics（ICB），2015：31-37.
[26] DAUGMAN J.How iris recognition works[J].IEEE Transactions on Circuits & Systems for Video Technology，2004，14（1）：21-30.
[27] JI T L，SUNDARESHAN M K，ROEHRIG H.Adaptive image contrast enhancement based on human visual properties[J].IEEE Transaction on Med imaging，1994，13（4）：573-586.
[28] LOSHCHILOV I，HUTTER F.SGDR：Stochastic gradient descent?with warm restarts[J].arXiv：1608.03983，2016.
[29] 陈彬.特殊场景下图像分类算法研究[D].北京：北京邮电大学，2020.
CHEN B.Reserch on image classification algorithm under special scenarios[D].Beijing：Beijing University of Posts and Telecommunications，2020.
[30] DAUGMAN J.The importance of being random：statistical principles of iris recognition[J].Pattern Recognition，2003，36（2）：279-291.
[31] SELVARAJU R R，COGSWELL M，DAS A，et al.Grad-CAM：Visual explanations from deep networks via gradient-based localization[J].International Journal of Computer Vision，2020，128（2）：336-359.