毕业论文外文翻译-指纹识别和验证的匹配系统.doc

《毕业论文外文翻译-指纹识别和验证的匹配系统.doc》由会员分享，可在线阅读，更多相关《毕业论文外文翻译-指纹识别和验证的匹配系统.doc（21页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、外 文 翻 译毕业设计题目：指纹识别与研究原文：Fingerprint Identification and Verification System using Minutiae Matching译文：指纹识别和验证的匹配系统原文：Fingerprint Identification and Verification System using Minutiae MatchingAbstract: Fingerprints are the most widely used biometric feature for person identification and verification i

2、n the field of biometric identification. Fingerprints possess two main types of features that are used for automatic fingerprint identification and verification: (i) global ridge and furrow structure that forms a special pattern in the central region of the fingerprint and (ii) Minutiae details asso

3、ciated with the local ridge and furrow structure.This paper presents the implementation of a minutiae based approach to fingerprint identification and verification and serves as a review of the different techniques used in various steps in the development of minutiae based Automatic Fingerprint Iden

4、tification Systems (AFIS). The technique conferred in this paper is based on the extraction of minutiae from the thinned, binarized and segmented version of a fingerprint image. The system uses fingerprint classification for indexing during fingerprint matching which greatly enhances the performance

5、 of the matching algorithm. Good results (92% accuracy) were obtained using the FVC2000 fingerprint databases.1. INTRODUCTION Fingerprints have been in use for biometric recognition since long because of their high acceptability, immutability and individuality. Immutability refers to the persistence

6、 of the fingerprints over time whereas individuality is related to the uniqueness of ridge details across individuals. The probability that two fingerprints are alike is 1 in 1.9 x 10151. These features make the use of fingerprints extremely effective in areas where the provision of a high degree of

7、 security is an issue. The major steps involved in automated fingerprint recognition include a) Fingerprint Acquisition, b) Fingerprint Segmentation, c) Fingerprint Image Enhancement, d) Feature Extraction e) Minutiae Matching, f) Fingerprint Classification. Fingerprint acquisition can either be off

8、line (inked) or Online (Live scan). In the inked method an imprint of an inked finger is first obtained on a paper, which is then scanned. This method usually produces images of very poor quality because of the non-uniform spread of ink and is therefore not exercised in online AFIS. For online finge

9、rprint image acquisition, capacitative or optical fingerprint scanners such as URU 4000, etc. are utilized which make use of techniques such as frustrated total internal reflection (FTIR)2, ultrasound total internal reflection3, sensing of differential capacitance4 and non contact 3D scanning5 for i

10、mage development. Live scan scanners offer much greater image quality, usually a resolution of 512 dpi, which results in superior reliability during matching in comparison to inked fingerprints. Segmentation refers to the separation of fingerprint area (foreground) from the image background 6. Segme

11、ntation is useful to avoid extraction of features in the noisy areas of fingerprints or the background. A Simple thresholding technique 7 proves to be ineffective because of the streaked nature of the fingerprint area. The presence of noise in a fingerprint image requires more vigorous techniques fo

12、r effective fingerprint segmentation. A good segmentation method should exhibit the following characteristics 8: It should be insensitive to image contrast It should detect smudged or noisy regions Segmentation results should be independent of whether the input image is an enhanced image or a raw im

13、age The segmentation results should be independent of image quality Ren et al. 8 proposed an algorithm for segmentation that employs feature dots, which are then used to obtain a close segmentation curve. The authors claim that their method surpasses directional field and orientation based methods 9

14、,10,11 for fingerprint image segmentation. Shen et al. 12 proposed a Gabor filter based method in which eight Gabor filters are convolved with each image block and the variance of the filter response is used both for fingerprint segmentation and quality specification. Xian et al. 13 proposed a segme

15、ntation algorithm that exploits a blocks cluster degree, mean and variance. An optimal linear classifier is used for classification with morphological post-processing to remove classification errors. Bazen et al. 14 proposed a pixel wise technique for segmentation involving a linear combination of t

16、hree feature vectors (i.e. gradient coherence, intensity mean and variance). A final morphological post-processing step is performed to eliminate holes in both the foreground and background. In spite of its high accuracy this algorithm has a very high computational complexity, which makes it impract

17、ical for real time processing. Klein et al.15 proposed an algorithm that employs HMMs to remove the problem of fragmented segmentation encountered during the use of different segmentation algorithms. For a good quality fingerprint feature extraction is much easier, efficient and reliable in comparis

18、on to a relatively lower quality fingerprint. The quality of fingerprints is degraded by skin conditions (e.g. wet or dry, cuts and bruises), sensor noise, non-uniform contact with sensor surface, and inherently low quality fingerprint images (e.g. those of elderly people, laborers). A significant p

19、ercentage of fingerprints are of poor quality, which must be enhanced for the recognition process to be effective. Thereare two major objectives of fingerprint enhancement i.e. i) to increase the contrast between ridges and valleys and ii) to connect broken ridges. These objectives can be fulfilled

20、by using a contextual filter whose characteristics vary according to the local context to be used for fingerprint enhancement instead of conventional image filters. The filter should posses the following characteristics:It should provide a low pass (averaging) effect alongthe ridge direction with th

21、e aim of linking small gaps and filling impurities due to pores or noise.It should have a band pass (differentiating) effect in the direction orthogonal to the ridges in order to increase the discrimination between ridges and valleys and top separate parallel linked ridges. Sherlock et al.16 propose

22、d an algorithm for fingerprint image enhancement that employs position-dependent Fourier-domain-filtering-based orientation smoothing and thresholding technique. Greenberg et al. 17 proposed two schemes for fingerprint enhancement. One method uses local histogram equalization, Wiener filtering, and

23、image binarization whereas the other method uses a unique anisotropic filter for direct grayscale enhancement. OGorman et al. 18,19 proposed a contextual filter based approach that utilizes four main parameters of fingerprint images at a given resolution i.e. maxima and minima of the ridge and valle

24、y widths to form a mother filter whose rotated versions are then convolved with the image to yield the enhanced output. Hong et al. 20 proposed an effective method based on Gabor filters for image enhancement. Gabor filters fulfill the requirements of a good fingerprint enhancement filter mentioned

25、earlier as they possess both the differentiating and averaging effects. Slight modifications to this technique were made by Greenberg et al. 21. Jiangwei et al. 22 modified the technique given by 20. Their approach models the sinusoidal shape of the ridge-valley structure in a closer manner and give

26、s better results. Tico et al. 23 made use of a ridge detecting technique based on the second directional derivative of the image to carry out fingerprint enhancement. Watson et al. 24 multiplied the Fourier transform of each 32x32 block by its power spectrum raised to a power k to produce an efficie

27、nt technique that can find its use in online fingerprint recognition systems. Fingerprints possess two major types of features: special type of pattern formed by the ridge and furrow structures in the central region of the fingerprints and minutiae detailsassociated with local ridges and furrows. Mi

28、nutiae are minute details such as ridge endings (a point where a ridge ends abruptly) or a ridge bifurcation (where a ridge breaks up into two ridges). Minutiae are characterized by the position, direction and type (Ridge ending or bifurcation). The global features are used for fingerprint classific

29、ation into six major classes whereas the minutiae details are used for fingerprint based person identification. Fingerprint analysts utilize the minutiae information for fingerprint identification and it is the most established technique in the field of AFIS. The accuracy of the technique is depende

30、nt upon the precision in the extraction of minutiae. There are two major type of methods that are used for minutiae extraction i) binarization based methods and ii) direct gray scale methods. In binarization-based methods some information is lost during binarization that can degrade the performance

31、of the minutiae extractor. Direct gray scale methods overcome these problems but may be difficult to implement and time consuming to operate. The typical approach for a binarization-based method involves a priori enhancement, binarization and then thinning. Various binarization and thinning approach

32、es are discussed in 25, 26, 27, 28, 29. Once a binary skeleton has been obtained, a simple image scan allows the pixels corresponding to the minutiae to be detected by finding the crossing number. The minutiae obtained as a result of minutiae extraction need to be filtered in order to remove the fal

33、se minutiae introduced. Various minutiae filtering techniques are proposed in 30,31. During enrollment the minutiae set obtained from an individuals fingerprint is stored as a template for that subject. In the authentication module, the fingerprint given as input is compared or matched with the temp

34、lates to provide the decision of authentication. This process is known as minutiae matching. For minutiae matching to be effective the input fingerprint should be registered to the template fingerprint using the minutiae information of both the fingerprints. After registration the minutiae sets are

35、compared using the spatial distance, which must be smaller than a particular threshold for two minutiae to be declared as matched. A minutia matching is usually performed by using Hough transform 32,33 or by pre- alignment 34. Fingerprint classification is the process of classifying a given fingerpr

36、int into a number of classes such as left loop, right loop, whorl, arch and tented arch, double loop etc. Fingerprint classification is used for indexing in the fingerprint database and can greatly improve the performance of an AFIS. A typical fingerprint classification approach is based on the extr

37、action of fingerprint singular points (core and delta) and the implementation of a rule based classification system that takes its decision based on the number and placement of core and delta in a fingerprint. Different classification algorithms are proposed in 35,36,37. The rest of the paper is org

38、anized as follows: Section II gives details about the method used, section III exhibits the results with section IV giving the conclusions.II. IMPLEMENTATION OF THE AFIS This paper introduces a prototype automatic identity authentication system that is capable of authenticating the identity of an in

39、dividual, using fingerprints. The main components of the AFIS are1. Fingerprint database2. Fingerprint features database 3. Enrollment Module4. Authentication module. The fingerprint database stores the fingerprint images. for this purpose. The features extracted from these fingerprints are stored i

40、n the features database along with the person ID. The objective of the enrollment module is to admit a person using his/her ID and fingerprints into a fingerprints database after the process of feature extraction. These features form a template that is used to determine or verify the identity of the

41、 subject, formulating the process of authentication. The component of the AFIS used for authentication is referred to as the authentication module. Figure 1 illustrates the different steps involved in the development of the AFIS. The details of these steps are given in the following subsections. Fin

42、gerprint segmentation is an important part of a fingerprint identification and verification system. However the time spent in segmentation is also crucial. The algorithms presented in 13 and 14 work quite well in the extraction of the required region but these algorithms have very high computational

43、 cost.We have developed an efficient algorithm that works with acceptable performance and has a lower computational cost. This algorithm is based only on the block coherence of an image. Coherence gives us a measure of how well the gradients of the fingerprint image are pointing in the same directio

44、n. In a window of size WxW around a pixel, the coherence is defined as: （1） Where Gx and Gy are the local gradients along X and Y directions respectively. The resulting gradient coherence image is smoothed by the use of a Gaussian smoothing filter to give a coherence image C(x, y). The smoothed imag

45、e is then binarized to produce a segmentation mask CB. The binarization is carried out by global thresholding as （2）Fig. 1: Steps required for feature extractionA B C DA. Orientation Field, B. Ridge Frequency Image C. Enhanced Image D. Thinned ImageIV. Minutiae Extraction Minutiae extraction was car

46、ried out using the crossing number approach 39. Crossing number of pixel p is defined as half the sum of the differences between pairs of adjacent pixels defining the 8-neighborhood of p. Mathematically （3） Where p0 to p7 are the pixels belonging to an ordered sequence of pixels defining the 8-neigh

47、borhood of p and val (p) is the pixel value. Fig.2: cn (p)=2,cn (p)=3 and cn (p)=1 representing a nonminutiae region, a bifurcation and a ridge ending Crossing numbers 1 and 3 correspond to ridge endings and ridge bifurcations respectively. An intermediate ridge point has a crossing number of 2. The

48、 minutiae obtained from this algorithm must be filtered to preserve only the true minutiae. The different types of false minutiae introduced during minutiae extraction include spike,bridge, hole, break, Spur, Ladder, and Misclassified Border areas. (See figure 5) Fig. 3 Types of false minutiae A B C D E FA. Spike, B. Bridge, C. Hole, D.Break, E. Spur F. LadderThe number of minutiae in a given area is also limited therefore the m