Research Of Dual-Mode Recognition Algorithm Based On Finger Crease And Finger Vein

Multi-modal biometric identification technology is based on information fusiontechnology, integrates the data of two or more biological feature to make decision. Comparingwith single modal recognition system, it makes identification information between differentmodal fully combination and overcomes interferences brought by outside factors easily. Thispaper fully consider the knuckle crease and finger vein, the advantages and disadvantages ofthese two modes, finger knuckle creases with the characteristics of stable characteristics,insufficient amount of information; finger vein with characteristics of in vivo identification,internal features, non-contact, security, etc. Integrate of these two biological characteristicsfor the formation of dual-mode biometric systems. The recognition performance will be betterthan the single-mode biometric systems based on knuckle creases, or finger vein.This paper designed and realized the dual-mode biometric identification system, basedon finger crease and finger vein. The system includes three parts: finger crease module, fingervein module and fusion module of multi-modal system.In the finger crease recognition module, first, the finger crease image was collected bydual-mode image acquisition devices， second, the image of the knuckle creases waspre-processed，including changes in image format, target area extraction, image enhancement,segmentation denoising, thinning and other operations，then the feature image of knucklecreases picture was got，the last operation of the image ismatching and feature extraction，thus the single-mode finger crease recognition system was completed. In the article, themathod of the direction of the filter was used to enhance the image of the finger crease andthresholding separate the threshold, median filtering was used for image denoising, finally, inthe session of feature match, based on relative distance, the matching algorithm was used tocomplete the finger crease image matching and recognition.In the finger vein recognition module, using the algorithms of finger crease module tocomplete image target area extraction, using the traditional method of Niblack for the veinimage segmentation, using the median filter and the method based on connected area tocomplete de-noising task, and then the vein image after image thinning to lay the foundationfor feature extraction and recognition.In multi-modal fusion module, the theory and structure of multi-modal identificationsystem were introduced at first, Then the dual-mode biological characteristics identificationsystem with images of knuckle crease and vein of finger were composed. Image preprocessing were done respectively, the characteristics of the image will be superimposed tocomplete the integration of the data layer. Later the feature extraction was done to the fusedimage, using the matching algorithm of relative distance to complete the task match. Throughanalysis of the results of identification to verify the dual-mode recognition system’s overallperformance is better than a single model of the finger knuckle creases or finger veinrecognition system. Further more, we proposed a decision-level fusion method based onsecondary decision making algorithm. Our results also show that this algorithm is better thansingle-mode recognition systems.