摘要
针对目前基于手掌静脉的身份识别问题,分析了掌脉图像的纹理特性,提出一种利用改进的S变换能量特征的识别方法。首先对掌脉图像进行二维离散正交S变换(2D-DOST),将变换后的频域图像作为掌脉纹理特征的一种体现,即形成S变换能量曲面(STES);然后将不同掌脉能量曲面进行减法运算,得到能量差曲面,进一步计算曲面的标准差,并以此为依据对不同掌脉进行分类识别。在通用的接触式标准掌脉图库上测试的结果表明,本文方法正确识别率为99.892 5%,识别时间为0.017 3s;同时使用自主设计的采集仪在非接触环境下测试的结果表明,本文方法正确识别率可达99.480 6%,识别时间为0.020 2s。相比其他典型和流行算法,本文算法提高了手掌静脉识别系统的性能,具有可行性和有效性。
Aiming at the identification problem based on palm vein, the texture characteristics of the palm vein image are analyzed, and a recognition method using improved S-transform energy feature is pro- posed. Firstly, the 2D discrete orthonormal S-transform （2I）-DOST） is used for palm vein image, and the frequency domain image is obtained after the transform as a reflection of palm vein texture features, which is called S transform energy surface （STES）, then the energy difference surface is got from the subtraction operation between the different palm vein energy surfaces. Furthermore, the standard devia- tion can be calculated from it,which will be a basis for the classification of different palm vein images. Experimental tests based on the common standard contact palm vein database show that the correct rec- ognition rate of this method is 99. 8925% ,and the recognition time is 0.017 3 s. At the same time,the actual effect tests are carried out using a self designed acquisition instrument in contactless environment. The results show that the correct recognition rate can reach 99. 480 6 %, and the recognition time is 0. 020 2 s. These demonstrate that the proposed method has certain advantages compared with other typical and popular algorithms, which is feasible and effective for improving the performance of palm vein recog- nition system.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2015年第9期1776-1782,共7页
Journal of Optoelectronics·Laser
基金
辽宁省教育厅科学研究(L2014132)资助项目
关键词
手掌静脉身份识别
二维离散正交S变换(2D-DOST)
纹理特征
采集仪器
palm vein identity recognition
two-dimensinal discrete orthonormal S-transform （2D- DOST）
texture feature
acquisition instrument
作者简介
林森（1980-），男，辽宁沈阳人，博士，讲师，主要从事机器视觉检测、模式识别等方面的研究．E-mail：lin_sen6@126．com
