计算机应用 ›› 2018, Vol. 38 ›› Issue (8): 2205-2210.DOI: 10.11772/j.issn.1001-9081.2018010183

• 人工智能 • 上一篇    下一篇

掌纹掌脉图像超小波域融合识别算法

李新春1, 曹志强2, 林森1, 张春华2   

  1. 1. 辽宁工程技术大学 电子与信息工程学院, 辽宁 葫芦岛 125105;
    2. 辽宁工程技术大学 研究生院, 辽宁 葫芦岛 125105
  • 收稿日期:2018-01-22 修回日期:2018-03-11 出版日期:2018-08-10 发布日期:2018-08-11
  • 通讯作者: 曹志强
  • 作者简介:李新春(1963-),男,辽宁喀左人,高级工程师,主要研究方向:无线传感器网络、数字图像处理;曹志强(1988-),男,河南驻马店人,硕士研究生,主要研究方向:图像处理、模式识别;林森(1980-),男,辽宁沈阳人,讲师,博士,主要研究方向:图像处理、模式识别;张春华(1994-),女,辽宁朝阳人,硕士研究生,主要研究方向:图像处理、模式识别。
  • 基金资助:
    辽宁省教育厅科学研究一般项目(L2014132);辽宁省自然科学基金面上项目(2015020100)。

Palmprint and palmvein image fusion recognition algorithm based on super-wavelet domain

LI Xinchun1, CAO Zhiqiang2, LIN Sen1, ZHANG Chunhua2   

  1. 1. School of Electronics and Information Engineering, Liaoning Technical University, Huludao Liaoning 125105, China;
    2. Graduate School, Liaoning Technical University, Huludao Liaoning 125105, China
  • Received:2018-01-22 Revised:2018-03-11 Online:2018-08-10 Published:2018-08-11
  • Supported by:
    This work is partially supported by the Scientific Research General Project of Liaoning Provincial Department of Education (L2014132); the Natural Science Foundation of Liaoning Province (2015020100).

摘要: 针对单一生物特征识别技术易受外界各种因素影响,识别率和稳定性有待提高的问题,提出一种掌纹掌脉图像超小波域融合识别算法NSCT-NBP。首先,对掌纹掌脉图像利用非下采样Contourlet变换(NSCT)进行分解,将得到的低频和高频子图像分别利用区域能量和图像自相似原理进行融合;然后,对融合后的图像利用近邻二值模式(NBP)提取纹理特征,获得特征向量;最后,通过计算特征向量间的汉明距离比较融合图像间的近似程度来计算等误率(EER)。在PloyU图库及自建图库上进行实验,结果表明,NSCT-NBP算法可获得最低的EER,分别为0.72%和0.96%,识别时间仅为0.0530 s和0.0871 s,与当前最优的基于小波变换和Gabor滤波器的掌纹掌脉融合方法相比,在两个图库上EER分别降低了4%和36.8%。NSCT-NBP算法能够有效融合掌纹掌脉图像的纹理特征,具有良好的识别性能,并且掌纹掌脉特征的融合增强了识别系统的安全性。

关键词: 图像处理, 多生物特征识别, 非下采样Contourlet变换, 近邻二值模式, 汉明距离

Abstract: Single biometric identification technology can be easily affected by various external factors, thus the recognition rate and stability are poor. A palmprint and palmvein image fusion recognition algorithm based on super-wavelet domain, namely NSCT-NBP, was proposed. Firstly, palmprint and palmvein images were decomposed by using Non-Subsampled Contourlet Transform (NSCT), then the obtained low-frequency and high-frequency sub-images were respectively merged by using the regional energy and image self-similarity principle. Secondly, the texture features were extracted from the fused images by using Neighbor based Binary Pattern (NBP), thus the eigenvector was got. Finally, the similarity of the fused images was calculated by Hamming distance of the feature vectors, to get Equal Error Rate (EER). The experiments were conducted on PolyU and a self-built database, the experimental results show that the lowest EER of NSCT-NBP algorithm were 0.72% and 0.96%, the identification time were only 0.0530 s and 0.0871 s. Compared with the current best palmprint-palmvein fusion method based on wavelet transform and Gabor filter, the EER of the two databases were reduced by 4% and 36.8%, respectively. The NSCT-NBP algorithm can effectively fuse the texture features of the palmprint-palmvein images and has good recognition performance. The fusion of palmprint-palmvein features can enhance the security of the recognition system.

Key words: image processing, multi-biometrics, Non-Subsampled Contourlet Transform (NSCT), Neighbor based Binary Pattern (NBP), Hamming distance

中图分类号: