Technology Study On Inherent Features Of Finger Vein

As the second generation of biometric identification technology,the finger vein recognition technology can overcome many shortcomings of other technologies based on hand features,such as finger prints,palm prints,hand shapes,hand vein,and so forth.The main reason for this phenomenon is because some inherent features of finger vein are utilized for recognition.The finger vein recognition technology has broad market prospects,since it has higher recognition accuracy,higher security,and convenient performance,which make it meet the actual demands.This article focuses on the image preprocessing of finger vein,feature extraction and matching algorithms,and designed a complete system of finger vein recognition.Some steps in terms of image preprocessing are as followings:first of all,the acquired finger-vein image are transformed from RGB to gray,and the normalization operations are executed to reduce the storage space,and then,the full finger regions are extracted by using Otsu threshold method based on the column direction,while the small block noises are eliminated through connected domain areas,subsequently,in order to obtain more clear finger image,the contrast adaptive histogram equalization and the median filtering are combined to enhance the quality of the finger vein images,at the end,considering that the non-contact vein images may be translated and/or rotated during the acquiring process,an extraction method is proposed based on finger vein image rotation correction of the region of interest,followed by the size normalization.As a results,some comparative experiments are conducted respectively on complete finger-vein regions and finger-vein regions of interest,the results show that the algorithm we used to extract the region of interest has better robustness and a higher recognition rate.In the aspect of feature extraction,three kinds of feature extraction algorithms based on the segmentation of vein are raised.Some detailed descriptions of the three segmentation methods are presented,say,the repeated line tracking,the direction of valley shape detector,and the wide line detector.By analyzing the impact of the parameters of the three methods for the segmentation of finger vein,the optimal set of parameters that can achieve the best effects are obtained.Then the template matching algorithm is adopted for matching for each of the three algorithms.Through comparing the different recognition performances from the aspects of the operation speed,the accuracy,and the precision of identification verification etc.,it can be concluded that the wide line detector has the best performances among others.LBP as a texture description operator can effectively describe the image texture feature information.This article studies both the advantages and the existing problems of the LBP operator.Most of the improved LBP operators are more robust to noise and image rotation,with less model species and lower dimensionality of features,and able to completely represent the image information.This article adopts several kinds of typical improved LBP operators to extract finger vein features,uses the histogram,binary code Chi Square distance,or Hamming distance for matching,and analyzes the robustness of the algorithm.According to the characters of the improved LBP operator,finger vein images are divided to two groups for comparison experiments,the experimental results indicate that the extracted features have the better separability and classification effects,which are extracted by the weighted LBP operator based on Chi Square distance matching and the MB-CSLBP operator based on Hamming distance matching.Finally,the complete finger vein recognition system is designed to verify the performance of the above proposed algorithms.The experimental results show that these algorithms can not only extract features of low dimensionality,and run at a high speed,get better recognition rate,but also do not require training data and have good generalization.