基于贝叶斯的指静脉识别算法及其FPGA实现
|
摘要
针对指静脉识别应用中异源手指匹配计算量大的问题,提出一种基于贝叶斯概率模型指导初步粗匹配的指静脉快速识别算法。首先,使用局部动态阈值法分割静脉纹路特征,并对静脉纹路特征进行形态学处理得到单像素静脉骨架;其次,提取能够有效表征静脉纹路信息的方向匹配点,在此基础上对图像特征分块粗匹配,使用贝叶斯概率模型计算粗匹配概率分布;然后,使用概率分布指导初步粗匹配;最后,使用平均Hausdorff计算指静脉特征距离,得到静脉的匹配结果。在FPGA开发平台实现了图像预处理硬件加速模块,并使用nios ii软核实现图像的特征提取和匹配识别。实验结果表明,所提出的方法能够显著的提高指静脉匹配速度。
In order to solve the problem of large amount of calculation between different fingers in finger vein recognition,a fast finger-finger recognition algorithm based on Bayesian probability model to guide the initial matching is proposed in this paper.Firstly,the local dynamic threshold method is used to segment the vein pattern features,and the vein pattern features are processed to obtain the skeleton of finger vein.Secondly,the direction matching points which can effectively characterize the vein texture information are extracted.On this basis,the image features are matched by block,and the Bayesian probability model is used to calculate the rough matching probability distribution.Then,use the probability distribution to guide the preliminary rough match.Finally,the mean Hausdorff is used to compute the finger vein characteristic distance and get the results.Image preprocessing hardware acceleration module is implemented in FPGA development platform,and image feature extraction and matching recognition are realized by using nios ii.The experimental results show that the proposed algorithm has a significant improvement in the matching speed and accuracy.
In order to solve the problem of large amount of calculation between different fingers in finger vein recognition,a fast finger-finger recognition algorithm based on Bayesian probability model to guide the initial matching is proposed in this paper.Firstly,the local dynamic threshold method is used to segment the vein pattern features,and the vein pattern features are processed to obtain the skeleton of finger vein.Secondly,the direction matching points which can effectively characterize the vein texture information are extracted.On this basis,the image features are matched by block,and the Bayesian probability model is used to calculate the rough matching probability distribution.Then,use the probability distribution to guide the preliminary rough match.Finally,the mean Hausdorff is used to compute the finger vein characteristic distance and get the results.Image preprocessing hardware acceleration module is implemented in FPGA development platform,and image feature extraction and matching recognition are realized by using nios ii.The experimental results show that the proposed algorithm has a significant improvement in the matching speed and accuracy.
引文
[1]KUMAR A,ZHOU Y.Human identification using processing[J].A Publication of the IEEE Signal Processing Society,2012,21(4):2228-2244.
[2]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.
[3]YANG L,YANG G,YIN Y,et al.Finger vein recognition with anatomy structure analysis[J].IEEETransactions on Circuits&Systems for Video Technology,2018,28(8):1892-1905.
[4]肖宾杰.基于独立成分分析的指静脉识别研究[J].电子测量与仪器学报,2012,26(10):841-845.
[5]YANG W M,LI W F,LIAO Q M.Finger vein verification based on neighbor pattern coding[J].IEICETransactions on Information and Systems,2013,96(5):1227-1229.
[6]LEE H C,KANG B J,LEE E C,et al.Finger vein recognition using weighted local binary pattern code based on a support vector machine[J].Frontiers of Information Technology&Electronic Engineering,2010,11(7):514-524.
[7]CAO D,YANG J,SHI Y,et al.Structure feature extraction for finger-vein recognition[C].Iapr Asian Conference on Pattern Recognition.IEEE Computer Society,2013:567-571.
[8]王科俊,马慧.使用改进的方向滤波与修正的Hausdorff距离的指静脉识别方法[J].计算机辅助设计与图形学学报,2011,23(3):385-391.
[9]李小刚,沈雷,张严严,等.基于细线距离顺序统计量指静脉快速识别算法[J].杭州电子科技大学学报,2017,37(4):20-24.
[10]TAHERI S,YEARWOOD J,MAMMADOV M,et al.Attribute weighted naive Bayes classifier using a local optimization[J].Neural Computing&Applications,2014,24(5):995-1002.
[11]房丙午,黄志球,王勇,等.基于贝叶斯网络的复杂系统动态故障树定量分析方法[J].电子学报,2016,44(5):1234-1239.
[12]康乐,张群,李涛泳,等.基于贝叶斯学习的下视三维合成孔径雷达成像方法[J].光学学报,2017,37(6):93-102.
[13]姜瑞,陈晓怀,王汉斌,等.基于贝叶斯信息融合的测量不确定度评定与实时更新[J].计量学报,2017,38(1):123-126.
[14]DUBUISSON M P,JAIN A K.A modified Hausdorff distance for object matching[C].International Conference on Pattern Recognition,IEEE,2002:566-568.
[15]ASAARI M S M,SUANDI S A,ROSDI B A.Fusion of band limited phase only correlation and width centroid contour distance for finger based biometrics[J].Expert Systems with Applications,2014,41(7):3367-3382.
[2]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.
[3]YANG L,YANG G,YIN Y,et al.Finger vein recognition with anatomy structure analysis[J].IEEETransactions on Circuits&Systems for Video Technology,2018,28(8):1892-1905.
[4]肖宾杰.基于独立成分分析的指静脉识别研究[J].电子测量与仪器学报,2012,26(10):841-845.
[5]YANG W M,LI W F,LIAO Q M.Finger vein verification based on neighbor pattern coding[J].IEICETransactions on Information and Systems,2013,96(5):1227-1229.
[6]LEE H C,KANG B J,LEE E C,et al.Finger vein recognition using weighted local binary pattern code based on a support vector machine[J].Frontiers of Information Technology&Electronic Engineering,2010,11(7):514-524.
[7]CAO D,YANG J,SHI Y,et al.Structure feature extraction for finger-vein recognition[C].Iapr Asian Conference on Pattern Recognition.IEEE Computer Society,2013:567-571.
[8]王科俊,马慧.使用改进的方向滤波与修正的Hausdorff距离的指静脉识别方法[J].计算机辅助设计与图形学学报,2011,23(3):385-391.
[9]李小刚,沈雷,张严严,等.基于细线距离顺序统计量指静脉快速识别算法[J].杭州电子科技大学学报,2017,37(4):20-24.
[10]TAHERI S,YEARWOOD J,MAMMADOV M,et al.Attribute weighted naive Bayes classifier using a local optimization[J].Neural Computing&Applications,2014,24(5):995-1002.
[11]房丙午,黄志球,王勇,等.基于贝叶斯网络的复杂系统动态故障树定量分析方法[J].电子学报,2016,44(5):1234-1239.
[12]康乐,张群,李涛泳,等.基于贝叶斯学习的下视三维合成孔径雷达成像方法[J].光学学报,2017,37(6):93-102.
[13]姜瑞,陈晓怀,王汉斌,等.基于贝叶斯信息融合的测量不确定度评定与实时更新[J].计量学报,2017,38(1):123-126.
[14]DUBUISSON M P,JAIN A K.A modified Hausdorff distance for object matching[C].International Conference on Pattern Recognition,IEEE,2002:566-568.
[15]ASAARI M S M,SUANDI S A,ROSDI B A.Fusion of band limited phase only correlation and width centroid contour distance for finger based biometrics[J].Expert Systems with Applications,2014,41(7):3367-3382.