Font Size: a A A

The Study Of Fusion Technology Based On Fingerprint And Finger Vein Image

Posted on:2010-01-14Degree:MasterType:Thesis
Country:ChinaCandidate:J Z LiuFull Text:PDF
GTID:2178360272997122Subject:Circuits and Systems
Abstract/Summary:PDF Full Text Request
With the advent of the information age, information security and confidentiality attract widespread attention. The rich physiological characteristics make the biometric identity recognition technology become an important method in the field of personal authentication. There are some biological characteristics such as fingerprint, palm print, iris, face, etc., in addition to including some behavioral characteristics such as signature, sound, keystroke, etc. From the statistical data, the fingerprint identification is the most widely used technology and the biometric products holding the largest market share. However, because the fingerprint is more easily wore, and it is difficult to extract images that the fingers are too dry or too wet, and there has the possibility of being stolen, and then the vein recognition technology is walking into our society. Although fingerprint and finger vein as a part of human body have their own advantages, only relying on a single method or single biometric sometimes difficult to meet the actual needs. Multibiometrics is the automated recognition of individuals based on their biological or behavioral characteristics, and involves different characteristics or different identify method to constitute the fusion system. With the appropriate fusion algorithm, multibiometrics normally are better than the performance of unimodal biometric systems, and it has become an important biological identification of the contents and direction of development.The specific content of this thesis is as follows:1. Imaging principle of vein image and the structure of vein acquisition device. Vein recognition technology gets the vein image using the absorption character under a particular near-infrared of red blood cells in vein. When the near-infrared (NIR) whose wavelength range is between 0.72μm and 1.10μm irradiates on the collected site of the hand, the hemoglobin of the vein of this part compared to the other tissues under the skin can absorb more IR radiation, and it can show the vein structure well. The first step of studying vein recognition is to obtain vein image, and the quality of the vein image is directly related to the subsequent step, which is about the accuracy of vein extraction and matching algorithm. Therefore, the performance of acquisition device for the vein image is one key step for researching vein recognition. The vein images referred to this paper are collected by the CCD image acquisition device, and the size of one image is 176 * 144.2. The pre-processing of a finger vein image. The pre-processing of vein image is the emphasis of the contents of this paper, and it mainly includes five parts:1) The purpose of orienting the finger region is to dispose the pixel values of the background as 0, and to reduce the effect to the subsequent steps of the noise on the background; 2) About the issue of normalization, this paper studies the normalized gray-scale algorithm. There may have a great gray-scale difference between the collected vein images because of distinct lights, so it is necessary to normalize the collected gray-scale images. This paper tells the mean-variance normalized gray-scale method; 3) The noise on the vein images are mainly salt-and-pepper noise and gaussian noise. First of all, we introduce three filter methods for vein images which are smoothing filter, median filter and Gaussian filter. According to the characters of the vein images used in this paper, a median filter is used to remove the speckle noise in images firstly, and then a 2-D Gaussian low-pass filter with a standard deviationσ= 0.8 is used to control the impact of high-frequency noise to the vein images; 4) This paper compares four typical threshold methods, that are fixed threshold method, overall average method, Otsu method, and NiBlack method. From the results of these methods we can see that basically there can not get an ideal segmentation result, but the effect of these threshold methods is improved from a single threshold value method to the multi-threshold method. Thus, the image threshold method has been proposed; 5) thinning the target image is to deal with the pixel width of refinement after binarization to 1. This paper describes a common thinning algorithm which is conditional thinning method based on mathematical morphology.3. The post-processing of the finger vein images. About the vein extraction, this thesis adopts an extraction method based on detailed feature points. The extracted vein features are the crossing points and the endpoints which belong to geometrical features. At the same time, their numbers, the relative positions, curvature and length, etc. can also be extracted as the features, and then comparing with the data deposited into the database. In this paper, an extraction algorithm based on 8-neighbor detailed feature points is introduced. This method not only will not be influenced by image translation, rotation, scaling and other factors, but also will not be affected by the changes of blood vessel size, if the thinning algorithm and the repair algorithm as well are reasonable usefulness.4. Multibiometrics identification system and its implementation. This paper studies multibiometrics technology based on fingerprint and finger vein. And launch the integration at matching score level. This fusion method using the matching scores of fingerprint and finger vein, then adopting the suitable fusion algorithm integrates the scores. The multibiometrics system can consumedly improve the veracity and the security of the personal identification.
Keywords/Search Tags:Biometric, finger vein recognition, image threshold method, information fusion
PDF Full Text Request
Related items