Research On Bi-modal Hand Feature Recognition Algorithm Based On Hand Shape And Finger Knuckle Print

Hand multi-biometric recognition system occupies an important position in the field of biometric recognition.Compared with the most traditional fingerprint identification,knuckle print and hand shape is more stable,not easy to abrase,forge and pilfer,in the aspect of image acquisition,the requirement of acquisition equipment and environment are not high,and the non-contact acquisition method also greatly improves the users' satisfaction,therefore finger knuckle print and hand shape of single mode identification system has attracted extensive attention both at home and abroad.A large number of studies show that multi-biometric fusion can greatly improve the recognition rate,anti-attack and robustness of biometric recognition system.However,multi-feature fusion undoubtedly increases the computational complexity,feature dimension and computation time.Therefore,the bimodal feature recognition system proposed in this paper presents a new solution to this problem.The hand shape feature is stable in a period of time,which has strong anti-counterfeiting and anti-attack performance;the feature extraction algorithm is simple and the recognition speed is high;the finger knuckle print is rich in textures,just like fingerprints,which can remain stable for a long time and contain high individual differences.The fusion of the two features can achieve complementary advantages and improve the recognition accuracy,stability and anti-attack performance of the biometric identification system.The main research and innovation of this paper is as follows:(1)The reason of low recognition rate caused by non-contact hand shape image acquisition is analyzed.In the process of hand recognition,different finger opening degree will cause contour deformation,which further affects the results.To overcome this difficulty,the paper proposes a new method of hand shape recognition that is not affected by finger root contour deformation.The finger skeleton and finger geometric feature are extracted by an improved finger thinning method.This method can solve the problems as the rough of image,burr,long operation time etc.in the process of finger skeleton extraction;at the same time,the hand shape recognition method based on geometric feature recognition can get better recognition results in non-contact acquisition.(2)According to the texture distribution characteristics of finger knuckle print image,a finger knuckle print ROI extraction method based on negative gradient extremum statistical distribution is proposed.The method solves the problem that the projectionbased method can not extract ROI or extract inaccurate images for low-resolution images;meanwhile,the method has good adaptability to illumination change;and the ROI image library displays that all hand images can extract accurate and effective ROI.(3)For finger knuckle print image recognition,the paper puts forward a recognition method combining global and local features.PCA is adopted as the global feature with simple features and fast speed recognition.On the other hand,in order to extract finer texture features that reflect detail features,a multi-scale joint distribution LBP operator is proposed,which can improve the traditional multi-scale LBP operator without considering the gray correlation between different scales and the relationship between local pixels and gray distribution in the whole image.In the process of global and local fusion,a two-layer serial fusion strategy is proposed.Firstly,the range of sample library is reduced according to the result of global matching,and then the matching result is further determined by fine matching.The proposed method combines the advantages of high global coarse matching speed and high local fine matching precision,which not only obtain higher recognition rate,but also optimize the recognition speed.(4)A hand image acquisition device is designed,which can acquire hand shape and finger knuckle print images at the same time in one acquisition,and a full-hand image database is established.According to the proposed image preprocessing method,a finger knuckle print ROI database is established to verify the effectiveness of the proposed algorithm.Based on the single modal feature recognition of hand-shaped and finger-shaped features,an algorithm based on finger-shaped and hand-shaped features fusion is put forward.The two-level fusion strategy of decision-making level and fractional level fusion is adopted.Through the method of hierarchical matching,for the sample which can't get the decision-making results,in the second level of fractional fusion,the discriminant-making results can be obtained according to the matching score weighting.The proposed feature fusion method not only improves the recognition rate of the system,but also ensures the time performance of the system in large database image matching.Meanwhile,the two biological features can complement each other,which ensure the reliability of the biological feature recognition system.To sum up,two biological features are extracted by using all-hand images: hand shape and finger knuckle print.In the fusion stage of bimodal feature,by optimizing the classification decision of each classifier,the recognition accuracy and speed of large database are improved.The experimental results show that the proposed method is effective and feasible,which lays a solid foundation for research on all-hand multi-modal identification in the future.