Research On Fusion Multiple Directional Features For Palmprint Recognition

With the development of the digital society,information security issues have received extensive attention.More and more occasions require simple,fast and effective identification to protect people's public safety and privacy.As an important part of the identity recognition field,biometrics technology is considered to be one of the most effective identification solutions and is an indispensable and important technology in government departments,business areas and daily life.Palmprint recognition technology is an emerging biometric recognition technology developed in the past decade.Compared with face recognition,iris recognition and fingerprint recognition which are concerned widely,palmprint recognition has many advantages,such as having much more features,collecting more conveniently,acquiring equipment with lower price and being accepted by users more easily.It is easy to be accepted by users and has good anti-counterfeiting performance.With the development of palmprint recognition technology,how to exert the advantages of palmprint and construct a robust palmprint recognition system is an urgent problem to be solved in palmprint recognition research.The palmprint recognition method based on multi-information fusion is a robust palmprint recognition scheme,which can effectively solve the problem of low security existing in palmprint recognition,and becomes a breakthrough of palmprint recognition technology.Among of the many palmprint recognition methods,the directional feature-based method is robust to illumination and rotation,and has achieved excellent recognition results.In view of the rich line directional features existing in the palmprint,this paper has conducted in-depth research on the method based on directional features.The main works are as follows:(1)Evaluate the performance of the line-shape based filter to extract palm directional features.Filters commonly used to extract palmprint directional features are Gabor filter and Modified Finite Radon Transform(MFRAT)filter.In addition,Log Gabor filter,Gaussian filter and Steerable filter can also be used to extract the directional characteristics of the palmprint.Based on the Local Line Directional Pattern(LLDP),this paper systematically evaluates the performance of the five filters to extract palm directional features.(2)Research the palmprint recognition algorithm based on directional feature descriptor.The Local Line Directional Fusion Pattern(LLDFP)algorithm is proposed: Firstly,the Gabor filter of 12 directions are used to convolve with the palmprint image.Secondly,the convolution values of adjacent directions are weighted and fused,11 sets of combined values are obtained.Finally,according to the combined values,the relevant four convolution results are selected,and the direction subscripts corresponding to the maximum and minimum values are used for LLDFP coding.The experimental results show that the LLDFP algorithm can effectively distinguish the differences between heterogeneous palmprints and improve the accuracy of the palmprint recognition system.(3)Research the palmprint recognition method of multi-feature fusion.The Fusion Directional Representations(FDR)algorithm is proposed: Firstly,the Block Band Limit Phase-Only Correlation_Directioanl Representation(BBLPOC_DR)algorithm is used to extract the Global Representations(GR)and Local Representations(LR)of the palmprint.Secondly,the LLDP(or LLDFP)algorithm and the CompC(or SMCC)algorithm are used to extract the LR of the palmprint.Finally,the three feature extraction methods are verified separately,and fusion in the score level to form the FDR algorithm.The experimental results show that FDR has better recognition performance than the existing palmprint recognition algorithm.