| Hand vein recognition technology is a new research hot spot in the bio-metric identification technology recently, which mainly includes back of hand vein recognition, palm vein recognition technology and finger vein recognition technology which is a hot spot in vein recognition technology in recent years. In this paper, research of finger vein infrared imaging and recognition methods is done. The work mainly contains near-infrared finger vein acquisition device’s design, finger vein image enhancement algorithm based on guided image filtering, finger vein image’s skeleton thinning algorithms, feature extraction methods, the design of classifier and MATLAB simulation system realization. The main research content is summarized as follows:The design of finger vein acquisition device based on near infrared principle. The keys of the device design are the selection of light source, the distance of finger and light source, the choice of near infrared camera and how to avoid the interference of outside light. Through many times of experiment, in this paper, near infrared LED tube which use three LEDs that the wavelength is 850 nm, the scattering Angle is 60 °, the rated power is 0.1w are chosen as near-infrared light source, CMOS near infrared camera is chosen as camera to make a finger vein acquisition device which size is 10 cm* 10 cm* 18 cm. And in the device a baffle with a groove is installed to fix finger position. Through this device stable finger vein images are obtained.Finger vein image detail enhancement algorithm is based on guided image filtering. Guided image filtering is good at edge maintaining, while finger vein image didn’t obtain good enhancement through guided image filtering directly. In this paper, firstly, the finger vein image which contrast ratio is low is used to do one self-guided filtering, and then a detail enhancement module is made to obtain the first finger vein detail enhancement image;furthermore, the first finger vein detail enhancement image is used as the next guided filter’s input image and guided image to obtain the enhanced finger vein image which using the detail enhancement module once more. The enhanced finger vein image which vein information is enhanced and skin domain is smoothed is helpful to the next image processes such as segmentation and thinning.Research of finger vein skeleton thinning algorithm. In order to solve the problems of the existing thinning algorithms, such as bifurcate problem, redundancy problem, the problem that the structure thinned is not match with the original figure structure, the improved EPTA parallel thinning algorithm (IEPTA) is proposed in this paper. The two son processes of ZS(ZHANG and SUEN) algorithm and the counters(the other two symmetrical son iterative processes) are added in IEPTA algorithm to achieve the skeleton which is closer to the center line of the image firstly. Then, in order to avoid the problem of the 2-pixel width slash thinning distortion, decision conditions are increased and 12 remove templates are designed under the condition A(P)= 2 to realize the skeleton. The experiments show that IEPTA algorithm can achieve 1-pixel skeleton which is match with the original finger vein image.The research of feature extraction method based on geometric feature. The traditional extraction method based on geometric feature which feature vector made by endpoints and branch points and the relative position of the feature points is not stable and cannot get high recognition rate. A new feature extraction method based on multi-pixel width skeleton structure feature is proposed in this paper, through designing classifier based on similarity coefficient, a recognition system based on multi-pixel skeleton structure feature extraction which recognition rate as high as 99.06%, recognition time is about 0.8s, is satisfied with actual application. The database is from the finger vein database(64 fingers, every finger has 15 samples) public by Tianjin Key Laboratory of intelligent signal and image processing. |
PDF Full Text Request