正在加载图片...
·970· 工程科学学报,第37卷,第7期 表5图像旋转处理时不同方法的识别率 Mechatronics and Automation.Luoyang,2006:1790 Table 5 Recognition rates with different approaches when the image is 2] Wang L,Leedham C.Cho S Y.Infrared imaging of hand vein rotated patterns for biometrie purposes.IET Comput Vision,2007,1(3- 方法 识别率/% 4):113 B3]Wu X Q,Gao E Y,Tang Y B,et al.A novel biometric system LBP 79.27 based on hand vein Proceedings of 5th International Conference LBpri 72.32 on Frontier of Computer Science and Technology.Changchun, CS-BP 87.91 2010:522 MB-BP 92.60 4]Yuksel A,AKarun L,Sankur B.Biometric identification through MB-CSLBP 97.15 hand vein patters /Proceedings of 2010 IEEE 18th Signal Pro- cessing and Communications Applications Conference.Diyarbakir, 第2~6组实验都进行了8×8分块处理.实验结 2010:708 [5]Yan Q Y.Study on Hand Vein Identification Algorithms [Disserta- 果证明,MB-CSLBP识别率最高,优于原始LBP及各种 tion].Beijing:North China University of Technology,2012 改进的LBP.原始LBP和LBP仅考量中心像素点与 (燕青字.手背静脉身份识别算法研究[学位论文].北京:北 其邻域的关系,具有不稳定性,所以在实验中两者的识 方工业大学,2012) 别率都相对较低.CS-LBP采用了中心对称的方法,使 6 Ojala T,Pietikainen M,Harwood D.Comparative study of texture 得对噪声更加不敏感.MB-BP利用局部范围的平均 measures with classification based on feature distributions.Pattern 值来代替单个像素之间比较的方法,能够提取更丰富 Recognit,1996,29(1):51 的局部信息,大尺度的特征对图像噪声的影响也更加 Ojala T,Pietikainen M,Maenpaa T.Multiresolution gray-scale and rotation invariant texture classification with local binary pat- 不敏感.所以CS-BP和MB-BP的识别率都相对较 terns.IEEE Trans Pattern Anal Mach Intell,2002,24(7):971 高.MB-CSLBP结合CS-BP和MB-BP的优势,利用 [8]Luo Y,Wu C M,Zhang Y.Facial expression feature extraction u- 优势互补,采用融合局部宏观特征和微观特征得到的 sing hybrid PCA and LBP.China Unis Posts Telecommun, 总特征,能够更准确地描述图像的整体信息,对噪声和 2013,20(2):120 光照也有更好的抑制作用.即使图像质量较差,MB- 9]Zhu Q,Wang Y W,Li C S.Visible light texture image classifica- CSLBP也能较好地提取出表征图像信息的特征,有较 tion using Gabor and IBP feature.J Comput Inf Syst,2013.9 (21):8415 强的鲁棒性 [0]Liu W F,Wang YJ,LiS J.LBP feature extraction for facial ex- 4结论 pression recognition.J Inf Comput Sci,2011,8(3):412 [11]Gao ZS,Yuan H Z,Yang J.Face recognition using fusion of 针对手背静脉图像微小纹理特征和宏观结构特征 Cartesian differential invariant and LBP.J Optoelectron Laser, 兼而有之的特点,本文通过结合MB-LBP和CS-LBP各 2010,21(1):112 自的优点,提出一种基于MB-CSLBP的手背静脉身份 12] Bereta M,Karczmarek P,Pedryez W,et al.Local descriptors in 识别方法.MB-CSLBP描述算子由于采用大区域取均 application to the aging problem in face recognition.Pattern Recognit,2013,46(10):2634 值算法,大尺度的描述能力,使得对图像噪声更加不敏 3] Yang X Y,Wu Q,Zhou Q.Method study for facial character 感.这种算子分别从宏观和微观角度对像素进行描 recognition based on improved LBP.J Inf Comput Sci,2013,10 述,从而使两种特征信息互补,更好地突出图像的整体 (9):2519 特性.MB-CSLBP采用中心对称的方式进行编码,使得 41 Heikkila M,Pietikainen M,Schmid C.Description of interest 对像素点的邻域关系描述性更强.中心对称方式编码 regions with local binary pattems.Pattern Recognit,2009,42 (3):425 后,获得更小的特征维数,因此也降低了运算量和存储 05] Liao S C.Zhu XX,Lei Z,et al.Learning multi-seale block lo- 量.MB-CSLBP与原始LBP及多种改进的LBP方法进 cal binary patters for face recognition /Proceedings of Adrances 行实验分析比较,结果表明,MB-CSLBP能更好地提取 in Biometrics-International Conference.Seoul,2007:828 手背静脉图像的特征信息,基于局部宏观特征和微观 06] Wang Y D,Li K F,Cui JL,et al.Study of hand-dorsa vein 特征结合的手背静脉身份识别方法有较好的识别率。 recognition Adeanced Intelligent Computing Theories and Appli- cations -6th International Conference on Intelligent Computing Changsha,2010:490 参考文献 [17]Wang Y D,Li K F,Shark L K,et al.Hand-dorsa vein recogni- [1]Wang K J,Zhang Y,Yuan Z,et al.Hand vein recognition based tion based on coded and weighted partition local binary patterns on multi supplemental features of multi-classifier fusion decision Proceedings of 2011 International Conference on Hand-Based /Proceedings of the 2006 IEEE International Conference on Biometrics.Hong Kong,2011:253工程科学学报,第 37 卷,第 7 期 表 5 图像旋转处理时不同方法的识别率 Table 5 Recognition rates with different approaches when the image is rotated 方法 识别率/% LBP 79. 27 LBPriu2 72. 32 CS-LBP 87. 91 MB-LBP9 92. 60 MB-CSLBP9 97. 15 第 2 ~ 6 组实验都进行了 8 × 8 分块处理. 实验结 果证明,MB-CSLBP 识别率最高,优于原始 LBP 及各种 改进的 LBP. 原始 LBP 和 LBPriu2仅考量中心像素点与 其邻域的关系,具有不稳定性,所以在实验中两者的识 别率都相对较低. CS-LBP 采用了中心对称的方法,使 得对噪声更加不敏感. MB-LBP 利用局部范围的平均 值来代替单个像素之间比较的方法,能够提取更丰富 的局部信息,大尺度的特征对图像噪声的影响也更加 不敏感. 所以 CS-LBP 和 MB-LBP 的识别率都相对较 高. MB-CSLBP 结合 CS-LBP 和 MB-LBP 的优势,利用 优势互补,采用融合局部宏观特征和微观特征得到的 总特征,能够更准确地描述图像的整体信息,对噪声和 光照也有更好的抑制作用. 即使图像质量较差,MB￾CSLBP 也能较好地提取出表征图像信息的特征,有较 强的鲁棒性. 4 结论 针对手背静脉图像微小纹理特征和宏观结构特征 兼而有之的特点,本文通过结合 MB-LBP 和 CS-LBP 各 自的优点,提出一种基于 MB-CSLBP 的手背静脉身份 识别方法. MB-CSLBP 描述算子由于采用大区域取均 值算法,大尺度的描述能力,使得对图像噪声更加不敏 感. 这种算子分别从宏观和微观角度对像素进行描 述,从而使两种特征信息互补,更好地突出图像的整体 特性. MB-CSLBP 采用中心对称的方式进行编码,使得 对像素点的邻域关系描述性更强. 中心对称方式编码 后,获得更小的特征维数,因此也降低了运算量和存储 量. MB-CSLBP 与原始 LBP 及多种改进的 LBP 方法进 行实验分析比较,结果表明,MB-CSLBP 能更好地提取 手背静脉图像的特征信息,基于局部宏观特征和微观 特征结合的手背静脉身份识别方法有较好的识别率. 参 考 文 献 [1] Wang K J,Zhang Y,Yuan Z,et al. Hand vein recognition based on multi supplemental features of multi-classifier fusion decision / / Proceedings of the 2006 IEEE International Conference on Mechatronics and Automation. Luoyang,2006: 1790 [2] Wang L,Leedham G,Cho S Y. Infrared imaging of hand vein patterns for biometric purposes. IET Comput Vision,2007,1( 3-- 4) : 113 [3] Wu X Q,Gao E Y,Tang Y B,et al. A novel biometric system based on hand vein / / Proceedings of 5th International Conference on Frontier of Computer Science and Technology. Changchun, 2010: 522 [4] Yüksel A,AKarun L,Sankur B. Biometric identification through hand vein patterns / / Proceedings of 2010 IEEE 18th Signal Pro￾cessing and Communications Applications Conference. Diyarbakir, 2010: 708 [5] Yan Q Y. Study on Hand Vein Identification Algorithms[Disserta￾tion]. Beijing: North China University of Technology,2012 ( 燕青宇. 手背静脉身份识别算法研究[学位论文]. 北京: 北 方工业大学,2012) [6] Ojala T,Pietikainen M,Harwood D. Comparative study of texture measures with classification based on feature distributions. Pattern Recognit,1996,29( 1) : 51 [7] Ojala T,Pietikainen M,Maenpaa T. Multiresolution gray-scale and rotation invariant texture classification with local binary pat￾terns. IEEE Trans Pattern Anal Mach Intell,2002,24( 7) : 971 [8] Luo Y,Wu C M,Zhang Y. Facial expression feature extraction u￾sing hybrid PCA and LBP. J China Univ Posts Telecommun, 2013,20( 2) : 120 [9] Zhu Q,Wang Y W,Li C S. Visible light texture image classifica￾tion using Gabor and LBP feature. J Comput Inf Syst,2013,9 ( 21) : 8415 [10] Liu W F,Wang Y J,Li S J. LBP feature extraction for facial ex￾pression recognition. J Inf Comput Sci,2011,8( 3) : 412 [11] Gao Z S,Yuan H Z,Yang J. Face recognition using fusion of Cartesian differential invariant and LBP. J Optoelectron Laser, 2010,21( 1) : 112 [12] Bereta M,Karczmarek P,Pedrycz W,et al. Local descriptors in application to the aging problem in face recognition. Pattern Recognit,2013,46( 10) : 2634 [13] Yang X Y,Wu Q,Zhou Q. Method study for facial character recognition based on improved LBP. J Inf Comput Sci,2013,10 ( 9) : 2519 [14] Heikkil M,Pietikinen M,Schmid C. Description of interest regions with local binary patterns. Pattern Recognit,2009,42 ( 3) : 425 [15] Liao S C,Zhu X X,Lei Z,et al. Learning multi-scale block lo￾cal binary patterns for face recognition / / Proceedings of Advances in Biometrics — International Conference. Seoul,2007: 828 [16] Wang Y D,Li K F,Cui J L,et al. Study of hand-dorsa vein recognition / / Advanced Intelligent Computing Theories and Appli￾cations — 6th International Conference on Intelligent Computing. Changsha,2010: 490 [17] Wang Y D,Li K F,Shark L K,et al. Hand-dorsa vein recogni￾tion based on coded and weighted partition local binary patterns / / Proceedings of 2011 International Conference on Hand-Based Biometrics. Hong Kong,2011: 253 · 079 ·
<<向上翻页
©2008-现在 cucdc.com 高等教育资讯网 版权所有