Research On Key Algorithms For Finger Vein Identification

Due to the higher security factor of biometric identification technology,its range of use has become wider and wider,and now it is gradually replacing traditional identification methods.There are many types of biometric recognition.The most commonly used are human face recognition and fingerprint recognition,but these several recognition methods have the possibility of being forged and have certain security risks.Finger vein recognition,due to its own characteristics,requires that the identified object be a living body,and the vein is under the skin surface,which is difficult to be forged.Therefore,finger vein recognition technology is more secure and has advantages over other recognition technologies.Relevant researches have been carried out on the problems of poor finger vein image quality,susceptibility to noise due to factors such as uneven illumination,and unsatisfactory recognition results due to inappropriate selection of feature extraction methods.The main research contents are as follows:(1)In order to make full use of the information of each sample and ensure the stability of the algorithm performance,this paper uses a sparse representation method with good robustness for finger vein image recognition.On the basis of the original sparse representation theory,considering the adverse effect of heterogeneous samples on sparse reconstruction,in this paper,the weight of each training sample of the sample is given so that the weight of training samples that are not related to the test sample is low;at the same time,the sparse representation The classification decision-making method of the classifier is improved,and the sum of the largest sparse coefficients in the class is used to replace the minimum reconstruction error,which greatly reduces the situation that the test sample belongs to multiple categories at the same time;(2)Combining the high-frequency and low-frequency features of the finger vein image,this paper proposes a method of finger vein recognition based on the doubledictionary cooperative sparse representation.This method uses the low-frequency and high-frequency features of the image so that the texture information and contour information in the image are obtained.Effective use,through experimental verification,can increase the proportion of correctly identified samples when the parameters are selected properly.However,the performance of the dual dictionary collaborative sparse representation method is more sensitive to the selection of image block size and wavelet basis function,and it is difficult to determine the most suitable block size and wavelet basis function for different images.(3)In view of the problem that it is difficult to accurately describe the finger vein image and the single sparse representation classifier with a single feature,the recognition effect of the single sparse representation classifier is poor.In this paper,based on the idea of multi-classifier decision fusion,a multi-sparse representation classifier model is constructed by combining multiple features Sparse means that the output information of the classifier determines the category to which the test sample ultimately belongs.In this paper,we vote based on a dictionary composed of each of the three features,and the final voting result determines the category of the test sample,that is,the finger vein recognition method based on decision fusion of multiple sparse classifiers.This method can avoid a feature that cannot accurately describe the image.Effective information,through the combination of multiple features,can effectively improve the accuracy of finger vein image recognition.The performance of this method is only sensitive to the weight of each sub-classifier,and the weight of each subclassifier can be obtained by training the training set without manual adjustment.The experimental results show that,even if the number of training samples is not large,this method can still get a better recognition effect.