第3期 胡娜,等:融合LBP纹理特征与B2DPCA技术的手指静脉识别方法 ·539· 训练样本数为5时,识别率达到96.75%;LBP+ method using rotation rectified finger vein images[J]. 2DPCA方法的识别率在3个、4个、5个训练样本 CAAI transactions on intelligent systems,2012,7(3): 时分别为95.75%、98.10%与96.75%,整体优于LBP十 230-234 PCA方法性能;而本文所提的LBP?,+B2DPCA [3]WU Zhendong,TIAN Longwei,LI Ping,et al.Generating stable biometric keys for flexible cloud computing authen- 方法的最优识别率达到了99.73%,性能十分优良。 tication using finger vein[J].Information sciences,2018. 100 433-434:431-447. 90 [4]YANG Jinfeng,SHI Yihu,JIA Guimin.Finger-vein image matching based on adaptive curve transformation[J].Pat- tern recognition,2017,66:34-43. 的 0 --LBP2 [5]QIU Shirong,LIU Yagin,ZHOU Yujia,et al.Finger-vein 50 。-LBP+PCA recognition based on dual-sliding window localization and --LBP2+2DPCA 40 LBP2+B2DPCA pseudo-elliptical transformer[J].Expert systems with ap- 30 3 4 5 plications,2016,64:618-632 训练样本个数 [6]ROSDI B A,SHING C W,SUANDI S A.Finger vein re- 图10 FV-USM静脉库采用不同主成分分析的识别率 cognition using local line binary pattern[J].Sensors,2011, Fig.10 Results obtained using the FV-USM image data- 11(12:11357-11371 bases by employing different reduction methods [7]HUI Ma,OLUWATOYIN P,SHULI S.A finger vein re- 实验结果表明,本文提出的方法无论在图像 cognition method using improved oriented filter and elast- 质量良好的天津静脉库还是存在部分图像质量较 ic registration[J.Research journal of applied sciences,en- 差的FV-USM静脉库的测试结果均达到了 gineering and technology,2013,6(7):1153-1159. 99%以上,识别性能较好,具备一定的实用性。 [8]LEE E C,PARK K R.Restoration method of skin scatter- ing blurred vein image for finger vein recognition[J].Elec- 5结束语 tronics letters,2009,45(21):1074-1076 [9]王贺.基于特征融合的手背静脉识别算法研究D1.吉林: 本文融合旋转统一的LBP算子与B2DPCA 吉林大学,2017. 技术对手指静脉图像进行有效的特征提取,改善 WANG He.Research on dorsal hand vein recognition al- 了在非接触式采集方式下的静脉图像存在的光照 gorithm based on feature fusion[D].Jilin:Jilin University, 不均因素导致识别率较低的问题。实验结果表 2017. 明,本文算法能大幅度提高识别率,特别是对于 [1O]杨文文,毛建旭,陈姜嘉旭.基于分块LBP和分块 对比度较差的FV-USM指静脉中的样本,相较于 PCA的指静脉识别方法[J].电子测量与仪器学报。 单一的LBP特征提取算法,传统的经典降维算法 2016.30(7):1000-1007. 和LBP与经典降维组合特征提取算法,拥有较好 YANG Wenwen,MAO Jianxu,CHEN Jiangjiaxu.Finger 的识别性能。由于融合了局部纹理特征算子与 vein recognition based on block LBP and block PCA[J]. B2DPCA技术,本文方法的识别时间比单独使用 Journal of electronic measurement and Instrumentation. 2016,30(7):1000-1007. 降维识别方法要长,而处理速度是衡量识别系统 [11]GUPTA P,GUPTA P.An accurate finger vein based veri- 性能的一个重要指标,因此在保持识别系统性能 fication system[J].Digital signal processing,2015,38: 的前提下如何进一步提高系统的识别速度是今后 43-52 研究的重点。 [12]KANG Wenxiong,CHEN Xiaopeng,WU Qiuxia.The 参考文献: biometric recognition on contactless multi-spectrum fin- ger images[J].Infrared physics and technology,2015,68: [1]WALUS M,BERNACKI K,KONOPACKI J.Impact of 19-27. NIR wavelength lighting in image acquisition on finger [13】王科俊,袁智.基于小波矩融合PCA变换的手指静脉 vein biometric system effectiveness[J].Opto-electronics 识别).模式识别与人工智能,2007,20(5:692-697. review,2017,25(4:263-268. WANG Kejun,YUAN Zhi.Finger vein recognition based [2]马慧,王科俊.采用旋转校正的指静脉图像感兴趣区域 on wavelet moment fused with PCA transform[J].Pattern 提取方法.智能系统学报,2012,7(3:230-234. recognition and artificial intelligence,2007,20(5): MA Hui,WANG Kejun.A region of interest extraction 692-697LBPu2 (8,1) LBPu2 (8,1) 训练样本数为 5 时,识别率达到 96.75%;LBP+ 2DPCA 方法的识别率在 3 个、4 个、5 个训练样本 时分别为 95.75%、98.10% 与 96.75%,整体优于 + PCA 方法性能;而本文所提的 +B2DPCA 方法的最优识别率达到了 99.73%,性能十分优良。 1 2 3 4 5 训练样本个数 30 40 50 60 70 80 90 100 正确识别率/% LBPu2 LBPu2+PCA LBPu2+2DPCA LBPu2+B2DPCA 图 10 FV-USM 静脉库采用不同主成分分析的识别率 Fig. 10 Results obtained using the FV-USM image databases by employing different reduction methods 实验结果表明,本文提出的方法无论在图像 质量良好的天津静脉库还是存在部分图像质量较 差 的 FV-US M 静脉库的测试结果均达到 了 99% 以上,识别性能较好,具备一定的实用性。 5 结束语 本文融合旋转统一的 LBP 算子与 B2DPCA 技术对手指静脉图像进行有效的特征提取,改善 了在非接触式采集方式下的静脉图像存在的光照 不均因素导致识别率较低的问题。实验结果表 明,本文算法能大幅度提高识别率,特别是对于 对比度较差的 FV-USM 指静脉中的样本,相较于 单一的 LBP 特征提取算法,传统的经典降维算法 和 LBP 与经典降维组合特征提取算法,拥有较好 的识别性能。由于融合了局部纹理特征算子与 B2DPCA 技术,本文方法的识别时间比单独使用 降维识别方法要长,而处理速度是衡量识别系统 性能的一个重要指标,因此在保持识别系统性能 的前提下如何进一步提高系统的识别速度是今后 研究的重点。 参考文献: WALUŚ M, BERNACKI K, KONOPACKI J. Impact of NIR wavelength lighting in image acquisition on finger vein biometric system effectiveness[J]. Opto-electronics review, 2017, 25(4): 263–268. [1] 马慧, 王科俊. 采用旋转校正的指静脉图像感兴趣区域 提取方法[J]. 智能系统学报, 2012, 7(3): 230–234. MA Hui, WANG Kejun. A region of interest extraction [2] method using rotation rectified finger vein images[J]. CAAI transactions on intelligent systems, 2012, 7(3): 230–234. WU Zhendong, TIAN Longwei, LI Ping, et al. Generating stable biometric keys for flexible cloud computing authentication using finger vein[J]. Information sciences, 2018, 433-434: 431–447. [3] YANG Jinfeng, SHI Yihu, JIA Guimin. Finger-vein image matching based on adaptive curve transformation[J]. Pattern recognition, 2017, 66: 34–43. [4] QIU Shirong, LIU Yaqin, ZHOU Yujia, et al. Finger-vein recognition based on dual-sliding window localization and pseudo-elliptical transformer[J]. Expert systems with applications, 2016, 64: 618–632. [5] ROSDI B A, SHING C W, SUANDI S A. Finger vein recognition using local line binary pattern[J]. Sensors, 2011, 11(12): 11357–11371. [6] HUI Ma, OLUWATOYIN P, SHULI S. A finger vein recognition method using improved oriented filter and elastic registration[J]. Research journal of applied sciences, engineering and technology, 2013, 6(7): 1153–1159. [7] LEE E C, PARK K R. Restoration method of skin scattering blurred vein image for finger vein recognition[J]. Electronics letters, 2009, 45(21): 1074–1076. [8] 王贺. 基于特征融合的手背静脉识别算法研究[D]. 吉林: 吉林大学, 2017. WANG He. Research on dorsal hand vein recognition algorithm based on feature fusion[D]. Jilin: Jilin University, 2017. [9] 杨文文, 毛建旭, 陈姜嘉旭. 基于分块 LBP 和分块 PCA 的指静脉识别方法[J]. 电子测量与仪器学报, 2016, 30(7): 1000–1007. YANG Wenwen, MAO Jianxu, CHEN Jiangjiaxu. Finger vein recognition based on block LBP and block PCA[J]. Journal of electronic measurement and Instrumentation, 2016, 30(7): 1000–1007. [10] GUPTA P, GUPTA P. An accurate finger vein based verification system[J]. Digital signal processing, 2015, 38: 43–52. [11] KANG Wenxiong, CHEN Xiaopeng, WU Qiuxia. The biometric recognition on contactless multi-spectrum finger images[J]. Infrared physics and technology, 2015, 68: 19–27. [12] 王科俊, 袁智. 基于小波矩融合 PCA 变换的手指静脉 识别[J]. 模式识别与人工智能, 2007, 20(5): 692–697. WANG Kejun, YUAN Zhi. Finger vein recognition based on wavelet moment fused with PCA transform[J]. Pattern recognition and artificial intelligence, 2007, 20(5): 692–697. [13] 第 3 期 胡娜,等:融合 LBP 纹理特征与 B2DPCA 技术的手指静脉识别方法 ·539·