北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 2017, Vol. 40 ›› Issue (5): 98-105.doi: 10.13190/j.jbupt.2017-077

• 研究报告 • 上一篇    下一篇

基于特定方向多尺度稀疏编码的手掌静脉特征识别方法

马欣, 景晓军   

  1. 北京邮电大学 信息与通信工程学院, 北京 100876
  • 收稿日期:2017-05-11 出版日期:2017-10-28 发布日期:2017-11-21
  • 作者简介:马欣(1981-),男,博士生,E-mail:xinma@bupt.edu.cn;景晓军(1965-),男,教授,博士生导师.

Palm Vein Recognition Method Based on Directional Multi-Scale Sparse Coding

MA Xin, JING Xiao-jun   

  1. School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China
  • Received:2017-05-11 Online:2017-10-28 Published:2017-11-21

摘要: 为提高手掌静脉特征识别的准确率和效率,提出了基于特定方向多尺度稀疏编码(DMSC)方法.首先计算全局方向和局部方向,然后利用多尺度Gabor对划分的子区域进行独立的多尺度方向滤波,并使用稀疏编码方法进行特征提取,利用外形编码产生特征矩阵,使用标准化汉明距离方法对特征矩阵进行相似性判断.实验结果表明,DMSC方法的类间区分性较好,识别力和效率高于现有的手掌静脉识别方法.

关键词: 稀疏编码, 掌静脉识别, 多尺度Gabor滤波

Abstract: A palm vein recognition method based on the directional multi-scale sparse coding (DMSC) was proposed to promote the recognition accuracy and operation efficiency. Firstly, the global and local orientation information were calculated from the palm vein region of interest area. Then, the directional multi-scale Gabor filter were convolved in every sub-region independently. Finally, a feature extraction mechanism based on spare coding was implemented, and the feature matrices were formed by appearance-based method. The similarity of two feature matrices were measured by normalized Hamming distance. Experiments show that the proposed DMSC method achieves higher recognition accuracy and operation efficiency than other palm vein recognition methods.

Key words: sparse coding, palm vein recognition, multi-scale Gabor filter

中图分类号: