What is a fingerprint [301460]

UNIVERSITATEA POLITEHNICA TIMIȘOARA

Facultatea de Electronică și Telecomunicații

Fingerprint Identification

Profesor Coordonator:

Georgiana Simion

Student: [anonimizat], where an minutiae points can be obtained by ridges characteristics where they are contained in ridge bifurcation and ending. A fingerprint identification is mainly applicable in investigations.

The skin on our palms and soles exhibits a flow-like pattern of ridges and valleys. [anonimizat], help the hand to grasp objects by increasing friction and improving the tactile sensing of surface textures. The “Friction Ridge Patterns” sidebar describes the nature and origin of these characteristics

Another important use of friction ridges is person identification. [anonimizat] a mark of identity. [anonimizat]. Superficial injuries such as cuts and bruises on the finger surface alter the pattern in the damaged region only temporarily; the ridge structure reappears after the injury heals.

One of the world’s [anonimizat] 1999. The IAFIS currently contains fingerprints of more than 60 [anonimizat], [anonimizat] 10-[anonimizat].

A fingerprint recognition system can be used for both verification and identification. [anonimizat] “enrolled” fingerprint of a specific user to determine if they are from the same finger (1:1 match). [anonimizat] a duplicate or false identity (1:N match). [anonimizat] a [anonimizat].

What is a fingerprint?

Fingerprints are ridge and valley patterns present on the surface of human fingertips. The purported uniqueness of fingerprints is characterized by three levels of features(see Fig. 1). [anonimizat], [anonimizat], are called level-1 features. Level-2 features mainly refer to minutia points in a local region; ridge endings and ridge bifurcations are the two most prominent types of minutiae.Level 3 features include all dimensional attributes at a [anonimizat], shape, [anonimizat], [anonimizat]. [anonimizat] (called minutiae) is regarded as the most distinctive feature and is most commonly used in fingerprint matching systems.

Fig. 1. Illustration of fingerprint features at three different levels. (a) A gray-scale fingerprint image, (b) level 1 features: orientation field and singular points (core shown as circle and delta shown as triangle), (c) level 2 features: ridge endings (red squares) and ridge bifurcations (blue circles) and (d) level 3 features: pores and dots.

Existing reconstruction algorithms essentially consist of two main steps: i) orientation field reconstruction and ii) ridge pattern reconstruction. The orientation field, which determines the ridge flow, can be reconstructed from minutiae and/or singular points. In the orientation field was reconstructed from the singular points (core and delta) using the zeropole model. However, the orientation field in fingerprints cannot simply be accounted for by singular points only. Cappelli proposed a variant of the zeropole model with additional degrees of freedom to fit the model to the minutiae directions. However, the orientation field reconstructed based on zero-pole model cannot be guaranteed when the singular points are not available. A set of minutiae triplets was proposed to reconstruct orientation field in triangles without using singular points. The algorithm proposed by Feng and Jain predicts an orientation value for each block by using the nearest minutia in each of the eight sectors.

The other step in fingerprint reconstruction is ridge pattern reconstruction based on the reconstructed orientation field. The ridge pattern reconstruction proposed only generates a partial skeleton of the fingerprint, which is obtained by drawing a sequence of splines passing through the minutiae. This method was further improved in by using linear integral convolution to impart texture-like appearance and low-pass filtering to get wider ridges. However, it can only generate a partial fingerprint, and the resulting ridge pattern is quite different from that of the target fingerprint

Due to its distinctiveness, compactness, and compatibility with features used by human fingerprint experts, minutiae-based representation has become the most widely adopted fingerprint representation scheme. But other representation schemes do show strong performance, i.e., Bioscrypt’s algorithm in FVC2002 and FVC2004 (Fingerprint Verification Competition). Some minutiae-based matching system also employ additional features, i.e., orientation field, singular points, ridge count, etc., to improve the matching accuracy. In these representation schemes, the grayscale image has the most information and features at all three levels are recorded (depending on the sensor); compared to grayscale image, phase image and skeleton image lose all Level 3 features and compared with phase image and skeleton image, the minutiae template further loses some Level 2 information, such as ridge path between minutiae. The widespread deployment of fingerprint recognition systems in various applications has caused concerns that compromised fingerprint templates may be used to make fake fingers, which could then be used to deceive all fingerprint systems the same person is enrolled in. Once compromised, the grayscale image is the most at risk. Leakage of a phase image or skeleton image is also dangerous since it is a trivial problem to reconstruct a grayscale fingerprint image from the phase image or the skeleton image. Fig. 2 shows the reconstructed grayscale image from the phase image Ψ(x,y) by cos(Ψ(x,y)) and that from the skeleton image by distance transform. In contrast to the above three representations, leakage of minutiae templates has been considered to be less serious as it is not trivial to reconstruct a grayscale image from the minutiae.

Fig. 2. Reconstruction of grayscale fingerprint image. (a) Reconstructed from phase image and (b) reconstructed from skeleton image.

Fingerprint Representation

Larkin and Fletcher, proposed representing a fingerprint image as a 2D amplitude and frequency modulated (AM-FM) signal:

(1)

which is composed of four components: the intensity offset , the amplitude , the phase , and the noise . Here, we are only interested in the phase , since ridges and minutiae are totally determined by the phase; the other three components just make the fingerprint appear realistic. Therefore, an ideal fingerprint can be represented as a 2D FM signal:

(2)

The gradient of the phase is also termed instantaneous frequency. In a fingerprint image, the direction of instantaneous frequency is normal to the local ridge orientation and the magnitude of instantaneous frequency is equal to the local ridge frequency. According to the Helmholtz Decomposition Theorem, the phase can be uniquely decomposed into two parts, the continuous phase and the spiral phase:

(3)

Hereinafter, the phase is also termed the composite phase to reflect the fact that it consists of the continuous phase and the spiral phase.

The continuous phase does not contain any rotational component and the integral of its gradient around any simple closed path is zero. For example, the continuous phase given by

(4)

corresponds to a grayscale image (cos) that looks like a whorl pattern (see Fig. 3). Its gradient (instantaneous frequency) is where is the angle in the polar coordinate system. The spiral phase consists of a set of N spirals (residues):

(5)

where and denote the coordinates of the nth spiral and denotes its polarity. A spiral with positive polarity is referred to as a positive spiral and with negative polarity is referred to as a negative spiral. The gradient of the spiral phase is not defined in the position of spirals. See Fig. 4 for the phase of a spiral and its gradient.

Fig. 3. Continuous phase for a whorl pattern. (a) Continuous phase given by (b) continuous phase modulo 2π, (c) grayscale image given by and (d) gradient of the continuous phase.

Fig. 4. (a) The spiral phase and (b) its gradient

Fingerprint Reconstruction

Given a set of N fingerprint minutiae, 1<n<N where and denote the location and direction of the nth minutia, respectively, the goal is to reconstruct the original fingerprint image modeled. In terms of the FM model, this input means that we are given 1) the spiral phase and 2) the direction of instantaneous frequency of the composite phase at the locations of the N minutiae. This is an ill-posed problem since the important information required to reconstruct the continuous phase of fingerprints, namely, the ridge frequency, is unknown. Information needed to reconstruct realistic fingerprints, such as brightness, contrast, the background noise of fingerprint sensor, and detailed ridge features (pores, ridge contours, etc.) is also not available. Thus, a more practical goal is to first estimate the FM representation of the original fingerprint, . The 8-bit grayscale fingerprint image is then computed as:

(6)

To obtain the phase , the following four steps are performed:

1. orientation field reconstruction,

2. estimation of gradient of continuous phase,

3. continuous phase reconstruction,

4. combination of the spiral phase and the continuous phase.

The flow chart of the proposed fingerprint reconstruction algorithm is depicted in Fig. 5.

Fig. 5. Flow chart of the proposed fingerprint reconstruction algorithm

1. Orientation Field Reconstruction

The image is divided into nonoverlapping blocks of 8×8 pixels and an orientation value is computed for each foreground block. A foreground mask for the fingerprint image is obtained by dilating the convex hull of minutiae using a disk-shaped structuring element of 8×8 pixels. The

local ridge orientation at block (m,n) is predicted by using the nearest minutia in each of the eight sectors. The minutia direction is doubled to make equivalent to . The cosine and sine components of of all of the K selected minutiae are summed as:

(7)

(8)

where is a weighting function. In our experiment, the reciprocal of the euclidean distance between the block center and the kth minutia is used in order to make minutiae direction dominate the ridge orientation of neighboring blocks. Other weighting functions, such as the Gaussian function, can also be used. Then, the orientation at block (m,n) is computed as:

) (9)

In the event that fingerprint singular points (core, delta) are also provided, a different approach is used to reconstruct the orientation field in order to avoid a possible shift of singularity .

2. Estimation of gradient of continuous phase

The gradient of the continuous phase at block can be computed as:

(10)

where and represent the gradients of the composite phase and of the spiral phase, respectively can be easily computed from the spiral phase in (5). Although is normal to local ridge orientation, its direction cannot be simply computed as +π/2 for two reasons. First, this may produce discontinuity in phase gradient since the orientation field is wrapped in the range [0,2π). Second, both ridge orientation and frequency

are not well defined in the neighborhood of minutiae. To deal with the first problem, we unwrap the initial orientation field, O , to obtain an unwrapped orientation field, . This is basically a phase unwrapping problem, except for the trivial difference that phase is

wrapped in the range [0,2π), while orientation is wrapped in the range [0,π). Starting from the top left-most foreground block, say block (m,n), whose initial orientation is directly set as its unwrapped orientation, the orientation at an adjacent block, say block (m+1,n), is unwrapped by adding kπ to its initial orientation (m+1,n). Here, k is an integer number that makes the following inequality hold:

(11)

3. Continuous Phase Reconstruction

The continuous phase of a fingerprint is modeled by piecewise planes at each foreground block (m,n) of 8×8 pixels:

(12)

where and denote the two components of and denotes the phase offset at block (m,n).

The only unknown value in (12), the phase offset , is estimated by the following algorithm. Starting with a queue containing the top left-most block (whose phase offset is assumed to be 0), in each iteration a block is obtained from the queue and each of its fourconnected neighbors is checked to see if it has been reconstructed (namely, the phase offset has been estimated). If one of the neighboring blocks has not been reconstructed, the phase offset of this block is estimated and it is put into the queue. This procedure is performed until the queue is empty (which means that the continuous phase has been reconstructed at all of the foreground blocks). An ancillary image is used to record the reconstructed blocks.

Fingerprint identification in Matlab Application

The quality of fingerprint pictures is the key to fingerprint identification, but in actual practice, the pictures we obtained often have all kinds of noise such as scars, perspiration, and stains, as well as some noise caused by uniform contact with fingerprint collecting devices.

Fingerprint identification technology mainly includes four functions: reading the fingerprint, feature extraction, saving the data and comparison.

Fig. 6 The flow of fingerprint identification

Image Segmentation with MATLAB

We get an original picture from the database of fingerprints.

a) From the fingerprint picture, we know there is a white area between the fingerprint picture and the region of background, so first we do the primary handling on the fingerprint picture, removing the outermost frame. The colors of the black background and white frame are very different, as well as their gray values, therefore, we can do the primary handling pictures on the basis of gray value.

b) Then we do more processing on the fingerprint picture, removing the rest of the background.

First, normalize the fingerprint picture that has been primary handled based on the formula:

(13)

If then normalize gray value to 255 and regarding the background here, and are the mean value and variance of expectation, respectively; according to the actual situation, and are the mean value and variance of fingerprint pictures.

Partitioning the fingerprint pictures into 8×8 small blocks, the variance of the gray value will be

smaller if it is a background area, and the variance of the foreground is larger, so we solve the

variance of each small block, and set one threshold.

Set the diamonds area where it is less than the threshold of the background area; we set the gray value as 255, while the area where it is more then threshold remains unchanged, so therefore, we can separate fingerprint pictures from background area.

2) Image Pretreatment with MATLAB

a) Image filtering, removing burrs, cavity process and process of binarization. Owing to the picture quality after segmentation, we need to do filtering, removing burrs, cavity process and binarization process with it, in order to make the fingerprint pictures clear,eliminating the unnecessary noises. It will be beneficial for more identification. First of all, we do the 3X3 median filter with the picture. The fingerprint picture after the median filter will go by binarization processing, becoming a binary image, and then the image is thinned. Because the depth is different according to the area of the fingerprint we obtained, if we do the binary segmentation for the rest of the picture based on the same threshold, it will cause a loss of a large amount information. In this paper, we need the thought of self-adaption threshold of binary locality, for each picture, the threshold we select should satisfy that the pixels that are greater than the threshold are equal to the pixels that are less than the threshold. Figure 7 shows the process of adaptive local threshold of binary:

In Figure 7, T is the mean of gray value of the fingerprint picture, Nh, and N1 are, respectively, the number of pixels that are equal or greater than T and the number of pixels that are equal or greater than T in the block (k,1), δ= w ×w ×10 %, and w is the size of the block (pixel). Then do more handling of the fingerprint picture in order to remove the burr and cavity in the picture.

b) Thin the image. Thinning is based on not influence connectedness of the streak line, and after image binarization, delete the edge pixels of the streak line until the streak line is as wide as a signal pixel.

3) Extracting the Detail Feature of the Fingerprint Detail Feature with MATLAB.

a) The Methods of Fingerprint Feature Extraction.

Fingerprint feature extraction is divided into two methods: one is extracting feature from the gray level image, and the other one is extracting feature from the binary image that has been thinned. In common, the algorithm that directly extracts feature from gray level image is tracking the gray fingerprint streak line, and on the basis of tracking results, we can find out the positions of features and judge the type of the features. This method eliminates the complicated fingerprint picture pretreatment process, but feature extraction algorithm is very complex. Because of the influence of noise and so on, feature information (position, and direction, etc.) are not accurate. Currently, most of the systems take advantage of the second method, which only needs a template of 3×3 and we can extract the endpoint and split point.

b) Extracting features points. The breakpoint and bifurcation (Fig. 8) are the principal characteristics of the thinned fingerprint image. In this paper, we construct the feature vector of the fingerprint using these principal characteristics, and the extracting method is the template matching method. The template matching method has the advantage of small calculation and quick speed.

Fig. 8 The breakpoint and bifurcation

In all states of the eight neighborhoods, there are eight characteristic conditions that satisfy the endpoint, and nine characteristic conditions that satisfy the bifurcation, which are shown in Figures 9 and 10, respectively.

Fig. 9 Template of break points

Fig. 10 Template of bifurcation

c) Eliminate the False Points in the Extracting Feature by mainly filtering features that don’t conform with the features of the fingerprint. False features have the following characteristics: most of them are on the edge; false features inside the image are close to each other, and two or more false features exist in very small area. According to these features, we propose two eliminating methods: the first method uses Image Slicer on the points that are on the edge, which cut off the points on the edge directly, or eliminate the features close to each other using the second method of Distance Threshold.

Conclusions

With the development of the technology, as a biological identification technique, fingerprint identification becomes perfect day after day, and people pay more and more attention to it. It is not only an important part of personal identifiers, but also has a wide application in each area of society, including police, army, customs, traffic, finance, and social insurance areas and departments.

BIBLIOGRAFIE

http://genetics.thetech.org/ask/ask68

http://www.nec.com/en/global/solutions/biometrics/technologies/fingerprint_identification.html

https://www.fbi.gov/about-us/cjis/fingerprints_biometrics/iafis/iafis

https://www.google.ro/webhp?sourceid=chrome-instant&ion=1&espv=2&es_th=1&ie=UTF-8#q=fingerprint%20identification&es_th=1

A Fingerprint Identification Technology Information Technology Essay

http://biometrics.cse.msu.edu/Publications/Fingerprint/JainFpMatching_IEEEComp10.pdf