基于胶囊网络的海量指纹纹型精准分类算法

doi:10.7523/j.issn.2095-6134.2020.03.012

摘要/Abstract

摘要： 指纹识别作为"物证之首"一直被认为是最可靠的生物特征识别方法，目前已经广泛应用在刑事侦查、居民身份认定及核实等领域。这类应用的特点在于需将待识别指纹与海量指纹数据库中的全部指纹做快速比对，以确定该枚指纹所有人的身份。为了提高指纹识别速度，海量指纹数据库会按照纹型拆分成若干类，待识别指纹仅和同类指纹做逐一比对。随着指纹采集相关的法律生效，近几年指纹数据库规模迅速扩大。一方面库内同类纹型的图像差异性显著增加，另一方面不同类指纹图像的相似性也在增加，指纹分类算法误分率大幅增加。"海量指纹纹型精准分类问题"迅速成为公安应用及指纹识别领域研究的热点。针对上述问题，提出一种基于胶囊网络的指纹纹型精准分类模型Cap-FingerNet。该模型一方面将胶囊网络独有的网络特性与指纹图像特有的自相似纹理特征相结合，可构成更为鲁棒的特征提取器及分类器。新引入的Batch-Normalization方法还可避免典型胶囊网络易于出现"梯度消失"的不足。另一方面，引入注意力胶囊机制，使得Cap-FingerNet较典型胶囊网络更准确且全面地提取出指纹图像细节信息，并使用全局压缩算法对胶囊进行有效挤压。Cap-FingerNet模型在某公安刑侦指纹数据库的4分类纹分类上获得99.63%的极高准确率，并在国际公开指纹数据集NIST-DB04的纹型4分类和5分类任务上分别测得96.25%和94.5%的准确率，取得了目前文献中最好成绩。

关键词: 指纹分类, 胶囊网络, 注意力胶囊, 全局压缩

Abstract: Fingerprint identification has been regarded as the most reliable biometric identification method, which has been widely used in the fields of criminal investigation, resident identification, and immigration personnel identification. The feature of these applications is that one query fingerprint needs to be quickly and accurately compared with all the fingerprints in the mass fingerprint database to identify the owner of the fingerprint. In recent years, the size of the fingerprint database was expanded rapidly with the law of fingerprint collection coming into effect. For one thing the image difference bewteen fingerprints in the same type increases significantly, for another the similarity of different types of fingerprint images also increases, which makes the misclassification rate of fingerprint classification algorithm greatly increase. At the same time, "massive fingerprint precision classification" has become a hot spot in the field of public security application and fingerprint identification. Aiming at solving the above problems, we propose an accurate fingerprint classification mode Caps-FingerNet based on capsule networks. On the one hand, the mode combines the unique network characteristics of the capsule networks with the individual self-similar texture features of fingerprint images to form a more robust feature extractor and classifier. Meanwhile, the batch-normalization method avoids the shortcoming of gradient disappearance in typical capsule networks. On the other hand, the attention capsule mechanism is innovatively introduced in this study, making Caps-FingerNet more accurate and comprehensive than typical capsule networks in extracting fingerprint image details, and the global squashing algorithm is used to squash the capsule effectively. Cap-FingerNet achieves an extreme accuracy of 99.63% in the four-class fingerprint classification of a provincial public security criminal fingerprint database. The accuracy rates of 96.25% and 94.5% were measured on the four and five classification tasks of NIST-DB04 fingerprint dataset, respectively.

Key words: fingerprint classification, capsule networks, attention capsule, global squashed

中图分类号:

TP391

李伯男, 赵彤, 吴敏. 基于胶囊网络的海量指纹纹型精准分类算法[J]. 中国科学院大学学报, 2020, 37(3): 387-397.

LI Bonan, ZHAO Tong, WU Min. Fast exact classification algorithm of massive fingerprint patterns based on capsule network[J]. , 2020, 37(3): 387-397.

参考文献

[1] Henry E R. Classification and uses of finger prints[M].3rd ed. London:HM Stationery Office, 1905.
[2] Cao K, Pang L, Liang J, et al. Fingerprint classification by a hierarchical classifier[J]. Pattern Recognition, 2013, 46(12):3186-3197.
[3] Jung H W, Lee J H. Noisy and incomplete fingerprint classification using local ridge distribution models[J]. Pattern Recognition, 2015, 48(2):473-484.
[4] Wang S, Zhang W W, Wang Y S. Fingerprint classification by directional fields[C]//Multimodal Interfaces. Pittsburgh:Fourth IEEE International Conference on Multimodal Interfaces, 2002:395-399.
[5] Yao Y, Frasconi P, Pontil M. Fingerprint classification with combinations of support vector machines[C]//International Conference on Audio-and Video-Based Biometric Person Authentication. Halmstad:Springer, Berlin, Heidelberg, 2001:253-258.
[6] Chang J H, Fan K C. A new model for fingerprint classification by ridge distribution sequences[J]. Pattern Recognition, 2002, 35(6):1209-1223.
[7] Hong L, Jain A. Classification of fingerprint images[C]//Proceedings of the Scandinavian Conference on Image Analysis. Kangerlussuaq:IAPR,1999, 2:665-672.
[8] Kawagoe M, Tojo A. Fingerprint pattern classification[J]. Pattern Recognition, 1984, 17(3):295-303.
[9] Zhang Q, Yan H. Fingerprint classification based on extraction and analysis of singularities and pseudo ridges[J]. Pattern Recognition, 2004, 37(11):2233-2243.
[10] Kamijo M. Classifying fingerprint images using neural network:Deriving the classification state[C]//IEEE International Conference on Neural Networks. San Francisco:IEEE,1993:1932-1937.
[11] Baldi P, Chauvin Y. Neural networks for fingerprint recognition[J]. Neural Computation, 1993, 5(3):402-418.
[12] Simonyan K, Vedaldi A, Zisserman A. Deep inside convolutional networks:visualising image classification models and saliency maps[J]. arXiv preprint arXiv:1312.6034, 2013.
[13] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014.
[14] Ren S, He K, Girshick R, et al. Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017(6):1137-1149.
[15] Girshick R. Fast r-cnn[C]//Proceedings of the IEEE International Conference on Computer Vision. Santiago:IEEE,2015:1440-1448.
[16] W Wen Y, Zhang K, Li Z, et al. A discriminative feature learning approach for deep face recognition[C]//European Conference on Computer Vision. Amsterdam:Springer, Cham, 2016:499-515.
[17] Schroff F, Kalenichenko D, Philbin J. Facenet:a unified embedding for face recognition and clustering[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Boston:IEEE,2015:815-823.
[18] Liu J, Deng Y, Bai T, et al. Targeting ultimate accuracy:face recognition via deep embedding[J]. arXiv preprint arXiv:1506.07310, 2015.
[19] Salamon J, Bello J P. Deep convolutional neural networks and data augmentation for environmental sound classification[J]. IEEE Signal Processing Letters, 2017, 24(3):279-283.
[20] Kim Y. Convolutional neural networks for sentence classification[J]. arXiv preprint arXiv:1408.5882, 2014.
[21] Ge S, Bai C, Liu Y, et al. Deep and discriminative feature learning for fingerprint classification[C]//2017 3rd IEEE International Conference on Computer and Communications (ICCC). Chengdu:IEEE, 2017:1942-1946.
[22] Wang R, Han C, Wu Y, et al. Fingerprint classification based on depth neural network[J]. arXiv preprint arXiv:1409.5188, 2014.
[23] Sabour S, Frosst N, Hinton G E. Dynamic routing between capsules[C]//Advances in Neural Information Processing Systems 30. Long Beach,2017:3856-3866.
[24] Hinton G E, Sabour S, Frosst N. Matrix capsules with EM routing[C]//International Conference of Learning Representation. Vancouver, 2018:3859-3869.
[25] Ioffe S, Szegedy C. Batch normalization:accelerating deep network training by reducing internal covariate shift[J]. arXiv preprint arXiv:1502.03167, 2015.
[26] Zeiler M D, Fergus R. Stochastic pooling for regularization of deep convolutional neural networks[J]. arXiv preprint arXiv:1301.3557, 2013.
[27] Everingham M, Van Gool L, Williams C K I, et al. The pascal visual object classes (voc) challenge[J]. International Journal of Computer Vision, 2010, 88(2):303-338.
[28] Deng L. The MNIST database of handwritten digit images for machine learning research[best of the web] [J]. IEEE Signal Processing Magazine, 2012, 29(6):141-142.
[29] Xiao H, Rasul K, Vollgraf R. Fashion-mnist:a novel image dataset for benchmarking machine learning algorithms[J]. arXiv preprint arXiv:1708.07747, 2017.
[30] LeCun Y, Huang F J, Bottou L. Learning methods for generic object recognition with invariance to pose and lighting[C]//Computer Vision and Pattern Recognition. Washington:IEEE, 2004, 2:II-104.
[31] Wilson C L. Massively parallel neural network recognition[C]//International Joint Conference on Neural Networks. IEEE, 1992, 3:227-232.
[32] Karu K, Jain A K. Fingerprint classification[J]. Pattern Recognition, 1996, 29(3):389-404.
[33] Senior A. A hidden Markov model fingerprint classifier[C]//Signals, Systems & Computers, 1997. Conference Record of the Thirty-First Asilomar Conference on. Beijing:IEEE, 1997, 1:306-310.
[34] Jain A K, Prabhakar S, Hong L. A multichannel approach to fingerprint classification[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1999, 21(4):348-359.
[35] Hong L, Jain A. Classification of fingerprint images[C]//Proceedings of the scandinavian conference on image analysis. Ystad:ACM,1999, 2:665-672.
[36] Cappelli R, Lumini A, Maio D, et al. Fingerprint classification by directional image partitioning[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1999, 21(5):402-421.
[37] Jain A K, Minut S. Hierarchical kernel fitting for fingerprint classification and alignment[C]//Object Recognition Supported by User Interaction for Service Robots. Quebec:IEEE, 2002, 2:469-473.
[38] Yao Y, Marcialis G L, Pontil M, et al. Combining flat and structured representations for fingerprint classification with recursive neural networks and support vector machines[J]. Pattern Recognition, 2003, 36(2):397-406.
[39] Chang J H, Fan K C. A new model for fingerprint classification by ridge distribution sequences[J]. Pattern Recognition, 2002, 35(6):1209-1223.
[40] Shalash W M, Abou-Chadi F. A fingerprint classification technique using multilayer SOM[C]//Proceedings of the Seventeenth National Radio Science Conference. Egypt:IEEE, 2000:C26/1-C26/8.