A Research Of Finger Vein Category Recognition Algorithm

Biometric technology is a technology that using the human physiologicalcharacteristics to achieve identification.With the high-speed development of the society’sinformatization、digitization and networked，various biometric systems are widly used inplaces such as banks、airports、military bases、R&D insituions.Because the finger veinrecognition technology with high anti-counterfeiting，high accuracy，easy to use and quicklyidentify shch prominent advantages，it become one of the most promising biometricrecognition technology in recent years.Research finger vein recognition technology has ahigh application value.This paper in-depth study on the category recognition algorithm in thefinger vein recognition technology，the main work includes:(1)Analyzed and summarized the finger vein classification recognition algorithm ofthis stage at home and abroad，and through the simulation experimental comparison theiradvantages and disadvantages.(2)According to the to the topological structural characteristics after the finger veinimage binarization，proposed an improved OPTA refinement algorithm.This algorithmimproved the inconsistency problem of the elimination template and the reservationtemplate，constructed eight limination templates and six reservation templates，and thenobtained the vein refined image of non-redundant dot’s single-pixel by cutting out the burrand repairing the image.(3)Proposed a finger vein recognition method based on Kernel Fisher DiscriminantAnalysis(KFDA).By introducing the kernel function，used dispersion matrix and divergencematrix as Fisher criterion， simulation of the finger vein database which waspretreatment.Experimental results showed this method with high recognition accuracy，itsprecision can reached98.2%.(4) Proposed a finger vein recognition method based on BP neural network.First，usedPrincipal Component Analysis(PCA) and Linear Discriminant Analysis(LDA) reducing thefinger vein image from a high-dimensional space to a low-dimensional space，removed theredundant information that contained in the image，extracted the eigenvector of the veinimage.Then used the extracted eigenvector as the inputs of BP neural network，training andrecognition. Experimental results showed this algorithm had a faster recognition speed andhigh recognition rate.Verified the effectiveness the algorithm in the finger vein recognition.