| Biometric identification technology mainly carries out identification through the physiological characteristics or behavioral characteristics of people.In recent years,finger vein recognition technology has become a hot spot for social research because of its high security and uniqueness.In this paper,we focus on several key issues of finger vein recognition technology,focusing on the pre-processing of finger vein image,vein feature extraction,matching algorithm and identity authentication of finger vein recognition model,etc.We conduct a more in-depth analysis and research.(1)Each process of finger vein image recognition algorithm is analyzed in detail,and each process of finger vein image recognition is experimentally processed and the corresponding experimental results are given.(2)A finger vein recognition algorithm that combines the LBP operator of rotation invariant mode and B2 DPCA technique is proposed for the situation that the traditional algorithm has a general effect on finger vein image recognition.First,the vein image is preprocessed and divided into several small image subblocks,and the texture features of each subblock,i.e.,LBP texture spectrum features,are extracted.Then,the feature vectors of all sub-blocks are dimensionally reduced using the bidirectional two-dimensional principal component analysis(B2DPCA)method to reduce the dimensionality of the feature vectors and improve the computational efficiency.Finally,the feature vectors of the vein images to be recognized are compared with the feature vectors of other samples and the Euclidean distance between them is calculated to perform the classification.(3)Finally,the recognition results were verified and analyzed by designing a finger vein recognition model.The finger vein image database was experimented with LBP,SIFT,SURF and the autonomous improved finger vein recognition algorithm,respectively.The results show that the recognition rates of traditional LBP,SIFT and SURF algorithms are97.7%,98.21% and 96.17% respectively,and the equal error rates are 4.46%,3.98% and6.06% respectively,while the recognition rate of the proposed finger vein recognition algorithm is 99.38% and the equal error rate is only 1.63%,which is much lower than LBP,SIFT and SURF algorithm,and it can meet the requirements of daily vein identification. |
