Research On Human Finger Vein Recognition Technology

The identity authentication and privacy protection are becoming more and more important in the information society. The personal identification technology based on biological characteristic has shown many advantages. It is convenient to using inherent physiology and behavior characteristics of human body. There are many types of biometric systems which are commercially available, such as fingerprints and faces used in safety inspection, entrance guard, encryption techniques, etc. However, these methods do not necessarily ensure confidentiality because the features are exposed on body surface which can be easily forged. Therefore, the patterns inside the human body are focused. The development of infrared imaging technique, pattern recognition and artificial intelligence makes personal identification based on human finger vein images possible. This paper proposes the overall developing paradigm of finger vein identification system via deep analysis of the hardware and software, and presents some effective techniques of finger vein image acquisition, vein feature extraction and matching. The main contents and innovations of the paper are as the follows:First, a systematic description about the technical backgrounds of infrared finger vein identification was presented. The difficulties of finger vein identification were introduced. The overall strategy and the plan of building an infrared finger vein identification system were proposed, which consisted of four parts including infrared finger vein image acquisition, image preprocess, vein line feature extraction and matching.Second, an infrared finger vein identification prototype device was designed, which composed of a PC and an on-line capture client. The capture client, composed of the light source module, filter/lens module, sensor module and communication interface module, was used to gather the infrared finger vein images. In the capture client, the dorsal side of the finger was illuminated by the infrared light, which could be auto-adjusted, and a CMOS camera with the infrared filter would capture the finger vein image. Because the fingers of humankind had different thickness, the system should capture a set of finger vein images under different light intensity in order to fuse the images into an ideal vein image.Third, the image fusion method to get an enhanced dynamic range image was proposed, which based on the camera response function and multi-intensity infrared finger vein images. The algorithm aimed to divide images into blocks and select best informed image blocks with regard to camera response curve and then fuse them up with a global fusion function to remove block discontinuity. Through this algorithm, an image with uniform brightness and contrast was obtained.Fourth, in vein feature extraction, the gray image morphology was used to detect the valley bottom of the gray profile vertical to the vein in the repeated line tracking. This method could smooth the profile and detect the valley bottom effectively. The result of the repeated line tracking was called the locus space as the finger vein feature extracted. In the post-process of the locus space, the method of image-resize and center-enhance was used to remove the flocky noise and to smooth the vein lines. The experiment result showed its availability.Fifth, a tri-value template fuzzy match algorithm was presented to reduce the effect of fuzzy edges and tips of the vein feature image. The proposed method would segment the vein feature image into three areas:subject area, fuzzy area and background area, and then compute the average distance of non-background point to non-background area as the dissimilarity score between the two templates. The proposed approach did not require knowledge of correspondence among those points in the two templates.Relative to the current state of finger vein recognition research, the theoretical analysis and techniques of vein image fusion, vein line extraction and vein feature matching in this thesis were creative.