Final year Project report

User Generated

Rorarmre

Computer Science

Description

Unformatted Attachment Preview

BIOMMETRIC TOUCH-LESS FINGERPRINT RECOGNITION Prepared by: Bankole Faroye Supervised by: Acknowledgement 1 2 Contents Abstract...................................................................................................... 5 Acknowledgement....................................................................................... 2 Chapter one: Introduction......................................................................... 6 1.1 INTRODUCTION ............................................................................... 7 1.2 Biometrics.............................................................................................. 8 1.3 Touch less Authentication Techniques................................................. 8 1.4 How Biometric Touch less Technologies Work.................................... 9 1.4.1 Preprocessing.................................................................................... 10 1.4.2 Extraction......................................................................................... 10 1.4.3 Verification & matching.................................................................. 10 1.4.4 Matches Are Based on Threshold Settings..................................... 11 1.5 Leading Biometric Technologies ...................................................... 12 1.6 Fingerprints as a Biometric............................................................... 13 1.6.1 Fingerprint Representation............................................................ 14 1.6.2 Minutiae.......................................................................................... 14 Chapter two: Motivation for the project............................................... 16 2.1 Problem Definition ........................................................................... 17 2.2 Motivation for the Project ................................................................ 18 2.3 About the Project............................................................................... 18 Chapter three: System design ................................................................ 20 2.1 System Level Design ......................................................................... 21 2.2 Algorithm Level Design..................................................................... 22 Chapter four: Fingerprint image preprocessing..................................... 24 4.1 Normalization.................................................................................... 25 4.2 Fingerprint segmentation.................................................................. 26 4.3 Fingerprint Extraction using STFT Analysis................................... 28 4.3.1 Ridge orientation image................................................................. 30 3 4.3.2 Ridge frequency image........................................................................ 32 4.3.3 Region mask................................................................................. 32 4.4 Core point Detection....................................................................... 34 Chapter five: Feature extraction ......................................................... 36 5.1 Minutia Extraction......................................................................... 37 5.2 Gabor filter..................................................................................... 39 Chapter six: Fingerprint verification................................................... 42 6.1 Classifiers-Binary........................................................................... 43 6.2 SVM- Support Vector Machine..................................................... 44 Chapter seven: Minutiae match .......................................................... 46 6.1 Alignment Stage ............................................................................ 48 6.2 Match Stage.................................................................................... 50 Chapter eight: System evaluation and conclusion .............................. 51 8.1 Evaluation of the system................................................................. 52 8.2 Conclusion....................................................................................... 54 Appendix............................................................................................... 55 REFERENCES ..................................................................................... 57 4 Abstract Touch-less fingerprint recognition system is a reliable alternative to conventional touch-based fingerprint recognition system. Touch-less system is different from conventional system in the sense that they make use of digital camera to acquire the fingerprint image where as conventional system uses live-acquisition techniques. The conventional fingerprint systems are simple but they suffer from various problems such as hygienic, maintenance and latent fingerprints. In this paper we present a review of touch-less fingerprint recognition systems that use digital camera. We present some challenging problems that occur while developing the touch-less system. These problems are low contrast between the ridge and the valley pattern on fingerprint image, non-uniform lighting, motion blurriness and defocus, due to less depth of field of digital camera. The touch-less fingerprint recognition system can be divided into three main modules: preprocessing, feature extraction and matching. Preprocessing is an important step prior to fingerprint feature extraction and matching. In this paper we put our more emphasis on preprocessing so that the drawbacks stated earlier can be removed. Further preprocessing is divided into four parts: first is normalization, second is fingerprint Segmentation, third is fingerprint enhancement and last is the core point detection. Feature extraction can be done by Gabor filter or by minutia extraction and the matching can be done by Support Vector Machine or Principal Component Analysis and three distance method. 5 CHAPTER ONE INTRODUCTION 6 1.1 INTRODUCTION Fingerprint recognition system is a biometric system that uses fingerprint as biometric input to this system. A fingerprint consists of patterns of ridges and valleys on the surface of a fingertip. Each individual has fingerprint which is different from the other. Actually this biometric system is a computer vision system which performs following functions: Image acquisition, Pre-processing, Feature Extraction, High-level processing or verification or matching. Basically, Fingerprint recognition system is an identification system that can be an Automated Fingerprint Identification System (AFIS) or a Non-automated Fingerprint Recognition System. Earlier, we used to take fingerprints using “ink techniques” in which black ink is spread on fingertip and it is pressed against a paper card, it is also called as “off-line fingerprint acquisition technique”. This technique is used in the law enforcement to acquire criminal’s fingerprints. Nowadays, live-scan acquisition technique is used in civil and criminal AFIS (Automated Fingerprint Identification system), that make use of sensors like optical, solidstate to acquire fingerprints. The proposed touch-less system can be an automated system or it can be a non-automated fingerprint recognition system. The basic difference between automated and non-automated system is that the AFIS was developed by the police department for identifying persons from over a large record of files where as non-automated fingerprint systems were developed for business purposes that replaced ID cards, passwords and other methods used for identifying. Conventional fingerprint systems are simple and require less processing as compared to the touch-less fingerprint recognition system, but they suffer from several problems like hygienic user interface, maintenance and latent fingerprints that are invisible from naked eyes and important for law enforcement agencies. 7 1.2 BIOMETRICS In the world of computer security, biometrics refers to authentication techniques that rely on measurable physiological and individual characteristics that can be automatically verified. In other words, we all have unique personal attributes that can be used for distinctive identification purposes, including a fingerprint, the pattern of a retina, and voice characteristics. Strong or twofactor authentication—identifying oneself by two of the three methods of something you know (for example, a password), have (for example, a swipe card), or is (for example, a fingerprint)— is becoming more of a genuine standard in secure computing environments. Some personal computers today can include a fingerprint scanner where you place your index finger to provide authentication. The computer analyzes your fingerprint to determine who you are and, based on your identity followed by a pass code or pass phrase, allows you different levels of access. Access levels can include the ability to open sensitive files, to use credit card information to make electronic purchases, and so on. 1.3 TOUCHLESS AUTHENTICATION TECHNIQUES While using the touch-based sensor interface, the fingertip needs to be placed over the interface so that a proper fingerprint image can be taken. But the touch-based sensors have several problems like the problem of contamination which occurs because of placing the fingertip over the same interface which is already used by other. This produces a low quality fingerprint image. Another problem is due to contact pressure, which creates physical distortions which are usually non-linear in arbitrary direction and strength. Moreover, the distortion occurs globally, while its deformation parameters could be different locally in a single fingerprint image 8 1.4 HOW BIOMETRIC TOUCH LESS TECHNOLOGIES WORK A touch-less fingerprint recognition system is based on remote sensing technology which allows us to take fingerprint images without any physical contact as in touch-based system. The touchless system makes use of touch-less fingerprint acquisition so as to capture the ridge-valley pattern which provides essential information for recognition. Touch less fingerprint recognition systems uses digital camera to acquire the fingerprint image. There is an advantage of using digital camera i.e. the fingerprint images captured with touch-less device are distortion free and present no deformation because these images are free from the pressure of contact. Fingerprint images captured with touch less system are distortion free and are most desirable to acquire minutiae in the same relative location and direction at every instance because this helps the authentication system to have low FAR (false acceptance rate) and FRR (false rejection rate). While there are strong advantages of using digital camera, there appear new weak points and they are; first, the contrast between the ridges and the valleys in fingerprint images obtained with a digital camera is low. Second, the depth of the field of the camera is small, thus some part of the fingerprint regions are in focus and some parts are out of focus. Third, the problem of motion blurriness in the acquired images. Thus, the main objective is to find solutions so as to overcome these drawbacks by putting main concern on the fingerprint image preprocessing. 9 Figure 1: shows the fingerprint image of one fingertip but with different minutiae because of physical pressure. Figure 2: Physical Distortion in touch-based sensors. The touch less fingerprint recognition system can be divided into three main modules and each module itself consists of some blocks. Each block of a module performs a special function over the input image. The three main sub-division or modules of Touch less Fingerprint Recognition System are: Pre-processing, Feature Extraction and verification or Matching. The block diagram of the touch-less fingerprint recognition system is shown in fig. 2. 10 1.4.1 PREPROCESSING Preprocessing is an important step prior to fingerprint feature extraction and matching. As the fingerprint images are captured using digital camera which had certain challenging problems as stated earlier so, these fingerprints require more preprocessing over them. Pre-processing is divided into four blocks. Normalization Fingerprint Segmentation Fingerprint Enhancement by STFT analysis Core Point Detection 1.4.2 EXTRACTION Most fingerprint identification methods use minutiae as the fingerprint features. The steps involved in minutiae extraction are smoothing, local ridge orientation estimation, ridge extraction, and thinning and minutiae detection. But for a small scale system, it is not efficient to process all the steps. So, we proposed Gabor filter-based feature extractor because its frequency and orientation representation are similar to those of human visual system. Also Gabor filter helps in smoothing out noise and preserving true ridge valley structures 1.4.3 VERIFICATION & MATCHING In verification systems, the step after extraction is to verify that a person is who he or she claims to be (i.e., the person who registered or enrolled). After the individual provides an identifier, the biometric is presented, which the biometric system captures, generating a test image that is based on the preprocessing. Fingerprint verification is done by SVM. SVM is a binary classifier which is based on the principle of structural risk minimization and maps an input sample to a high-dimensional feature space. SVM could optimally separate the two classes of genuine and imposters by constructing a hyper-plane. 11 1.4.4 MATCHES ARE BASED ON THRESHOLD SETTINGS No match is ever perfect in either verification or identification system, because every time a biometric is captured, the template is likely to be unique. Therefore, biometric systems can be configured to make a match or no-match decision, based on a predefined number, referred to as a threshold, which establishes the acceptable degree of similarity between the test image and the registered image. After the comparison, a score representing the degree of similarity is generated, and this score is compared to the threshold to make a match or no-match decision. Depending on the setting of the threshold in identification systems, sometimes several reference images can be considered matches to the test image, with the better scores corresponding to better matches. 1.5 LEADING BIOMETRIC TECHNOLOGIES A growing number of biometric technologies have been proposed over the past several years, but only in the past 5 years have the leading ones become more widely deployed. Some technologies are better suited to specific applications than others, and some are more acceptable to users. We describe seven leading biometric technologies: Facial Recognition Fingerprint Recognition Hand Geometry Iris Recognition Signature Recognition Speaker Recognition 12 Fingerprint Recognition Fingerprint recognition is one of the best known and most widely used biometric technologies. Automated systems have been commercially available since the early 1970s, and at the time of our study, we found there were more than 75 fingerprint recognition technology companies. Until recently, fingerprint recognition was used primarily in law enforcement applications. Fingerprint recognition technology extracts features from impressions made by the distinct ridges on the fingertips. The fingerprints can be either flat or rolled. A flat print captures only an impression of the central area between the fingertip and the first knuckle; a rolled print captures ridges on both sides of the finger. An image of the fingerprint is captured by a scanner, enhanced, and converted into a template. Scanner technologies can be optical, silicon, or ultrasound technologies. Ultrasound, while potentially the most accurate, has not been demonstrated in widespread use. In 2002, we found that optical scanners were the most commonly used. During enhancement, “noise” caused by such things as dirt, cuts, scars, and creases or dry, wet or worn fingerprints is reduced, and the definition of the ridges is enhanced. Approximately 80 percent of vendors base their algorithms on the extraction of minutiae points relating to breaks in the ridges of the fingertips. Other algorithms are based on extracting ridge patterns. 1.6 FINGERPRINTS AS A BIOMETRIC Among all biometric traits, fingerprints have one of the highest levels of reliability and have been extensively used by forensic experts in criminal investigations. A fingerprint refers to the flow of ridge patterns in the tip of the finger. The ridge flow exhibits anomalies in local regions of the fingertip (Figure), and it is the position and orientation of these anomalies that are used to represent and match fingerprints. 13 Although not scientifically established, fingerprints are believed to be unique across individuals, and across fingers of the same individual. Even identical twins having similar DNA, are believed to have different fingerprints. Traditionally, fingerprint patterns have been extracted by creating an inked impression of the fingertip on paper. The electronic era has ushered in a range of compact sensors that provide digital images of these patterns. These sensors can be easily incorporated into existing computer peripherals like the mouse or the keyboard (figure), thereby making this mode of identification a very attractive proposition. This has led to the increased use of automatic fingerprint-based authentication systems in both civilian and law enforcement applications. 14 1.6.1 FINGERPRINT REPRESENTATION The uniqueness of a fingerprint is determined by the topographic relief of its ridge structure and the presence of certain ridge anomalies termed as minutiae points. Typically, the global configuration defined by the ridge structure is used to determine the class of the fingerprint, while the distribution of minutiae points is used to match and establish the similarity between two fingerprints. Automatic fingerprint identification systems, that match a query print against a large database of prints (which can consist of millions of prints), rely on the pattern of ridges in the query image to narrow their search in the database (fingerprint indexing), and on the minutiae points to determine an exact match (fingerprint matching). The ridge flow pattern itself is rarely used for matching fingerprints. 1.6.2 MINUTIAE Minutiae, in fingerprinting terms, are the points of interest in a fingerprint, such as bifurcations (a ridge splitting into two) and ridge endings. Examples are: a.) ridge endings - a ridge that ends abruptly b.) ridge bifurcation - a single ridge that divides into two ridges c.) short ridges, island or independent ridge - a ridge that commences, travels a short distance and then ends d.) ridge enclosures - a single ridge that bifurcates and reunites shortly afterward to continue as a single ridge e.) spur - a bifurcation with a short ridge branching off a longer ridge f.) crossover or bridge - a short ridge that runs between two parallel ridges 15 Minutiae also refer to any small or otherwise incidental details. But the focus when matching is only on the 2 main minutiae; ridge ending and ridge bifurcation. 16 CHAPTER ONE MOTIVATION FOR THE PROJECT 17 2.1 PROBLEM DEFINITION We propose a simple and effective approach for Biometric fingerprint image enhancement and minutiae extraction based on the frequency and orientation of the local ridges and thereby extracting correct minutiae points. Automatic and reliable extraction of minutiae from fingerprint images is a critical step in fingerprint matching. The quality of input fingerprint images plays an important role in the performance of automatic identification and verification algorithms. In this project I presents a fast fingerprint enhancement and minutiae extraction algorithm which improves the clarity of the ridge and valley structures of the input fingerprint images based on the frequency and orientation of the local ridges and thereby extracting correct minutiae. 18 Touch less Fingerprint based identification has been one of the most successful biometric techniques used for personal identification. Each individual has unique fingerprints. A fingerprint is the pattern of ridges and valleys on the finger tip. The conventional fingerprint systems are simple but they suffer from various problems such as hygienic, maintenance and latent fingerprints. In this paper we present a review of touchless fingerprint recognition systems that use digital camera. We present some challenging problems that occur while developing the touch-less system. These problems are low contrast between the ridge and the valley pattern on fingerprint image, non-uniform lighting, motion blurriness and defocus, due to less depth of field of digital camera. Touch-less fingerprint recognition system is a reliable alternative to conventional touchbased fingerprint recognition system. Touch-less system is different from conventional system in the sense that they make use of digital camera to acquire the fingerprint image where as conventional system uses live-acquisition techniques. 2.2 MOTIVATION FOR THE PROJECT Accurate automatic personal identification is critical in wide range of application domains such as national ID cards, electronic commerce and automatic banking. Biometrics, which refers to automatic identification of a person based on his or her personal physiological or behavioral characteristics, is inherently more reliable and more capable in differentiating between a reliable person and a fraudulent impostor than traditional methods such as PIN and passwords. Touch less fingerprint identification is one of the most reliable biometric technology among the different major biometric technologies which are either currently available or under investigation. The objective of our project is to implement the image enhancement and minutiae extraction algorithm which is capable of doing the matching between different digitized 19 fingerprints of standard image file formats namely; BMP, JPEG with high level of accuracy and confidence. 2.3 ABOUT THE PROJECT We propose a simple and effective approach for fingerprint image enhancement and minutiae extraction based on the frequency and orientation of the local ridges and thereby extracting correct minutiae. Automatic and reliable extraction of minutiae from fingerprint images is a critical step in fingerprint matching. The quality of input fingerprint images plays an important role in the performance of automatic identification and verification algorithms. In this project we presents a touch less fingerprint enhancement and minutiae extraction algorithm which improves the clarity of the ridge and valley structures of the input fingerprint images based on the frequency and orientation of the local ridges and thereby extracting correct minutiae. Fingerprint based identification has been one of the most successful biometric techniques used for personal identification. Each individual has unique fingerprints. A fingerprint is the pattern of ridges and valleys on the finger tip. A fingerprint is thus defined by the uniqueness of the local ridge characteristics and their relationships. Minutiae points are these local ridge characteristics that occur either at a ridge ending or a ridge bifurcation. A ridge ending is defined as the point where the ridge ends abruptly and the ridge bifurcation is the point where the ridge splits into two or more branches. In conventional system Automatic minutiae detection becomes a difficult task in low quality fingerprint images where noise and contrast deficiency result in pixel configurations similar to that of minutiae. This is an important aspect that has been taken into consideration in this presentation for extraction of the minutiae with a minimum error in a particular location. A complete minutiae extraction scheme for automatic fingerprint recognition systems is presented. The proposed method uses improving alternatives for the image enhancement process, leading consequently to an increase of the reliability in the minutiae extraction task. 20 CHAPTER THREE SYSTEM DESIGN 21 3.1 SYSTEM LEVEL DESIGN A touch less fingerprint recognition system constitutes of fingerprint acquiring device, minutia extractor and minutia matcher. 22 Figure 3.1.1 Simplified Fingerprint Recognition System For fingerprint acquisition, optical or semi-conduct sensors are widely used. They have high efficiency and acceptable accuracy except for some cases that the user’s finger is too dirty or dry. However, the testing database for my project consists of scanned fingerprints using the ink and paper technique because this method introduces a high level of noise to the image and the goal of designing a recognition system is to work with the worst conditions to get the best results. The minutia extractor and minutia matcher modules are explained in detail later on in this paper. 3.2 ALGORITHM LEVEL DESIGN 23 To implement a minutia extractor, a three-stage approach is widely used by researchers. They are preprocessing, minutia extraction and verification stage. Fig 3.2.1Block diagram of touch less fingerprint analysis and comparison We propose a simple and effective approach for Touch less finger recognition and comparison based on the frequency and orientation of the local ridges and thereby extracting correct minutiae points. For the fingerprint image preprocessing stage, Histogram Equalization and Fourier Transform are used to do image enhancement. And then the fingerprint image is binarized using the locally adaptive threshold method. The image segmentation task is fulfilled by a three-step approach: block direction estimation by ridge orientation, segmentation by skin color detection, adaptive thresholding, morphological processing. Most fingerprint identification methods use minutiae as the fingerprint features. The steps involved in minutiae extraction are smoothing, local ridge orientation estimation, ridge extraction, and thinning and minutiae detection. But for a small scale system, it is not efficient to process all the steps. So, we proposed Gabor filter-based feature extractor because its frequency and orientation representation are similar to those of human visual system. Also Gabor filter helps in smoothing out noise and preserving true ridge valley structures. 24 CHAPTER FOUR FINGERPRINT IMAGE PREPROCESSING 25 4.1 NORMALIZATION Since fingerprint images can be captured at different instants of the day or on different days, the intensity for each image may exhibit variations. To avoid these light intensity variations, the test images are normalized so as to have an average intensity value with respect to the registered image. The average intensity value of the registered images is calculated as summation of all pixel values divided by the total number of pixels. Similarly, average intensity value of the test image is calculated. The normalization value is calculated as: Normalization is the first preprocessing operation. It can be done in two ways for two different purposes. In the first way, normalization is done so as to minimize the non-uniform lighting problem. It can be done by changing the dynamic range of the pixel intensity values. It calculates the mean and variance of an image and thus reduces the difference of the illumination. The normalization of is computed as follows: The estimation of the mean and standard deviation is performed through spatial smoothing. The other way, an input fingerprint image is normalized so that it has a pre-specified mean and variance . Normalization of an image is carried out after conversion of image from RGB to grayscale. The fingerprint images captured by using digital camera are in RGB format, so RGB to grayscale image conversion is done. A grayscale image has 256 different gray levels which are sufficient for the recognition of most natural objects .In a gray-level fingerprint image, ridges 26 and valleys in a local neighborhood form a sinusoidal-shaped plane wave which has a welldefined frequency and orientation. Now after conversion, the grayscale image is then given as an input to the normalization block where the image can be normalized. Let I(i,j) denote the graylevel value at pixel (i,j) , M and V denote the estimated mean and variance of I, respectively, and G(i,j) denote the normalized gray-level value at pixel (i,j). The normalized image is defined as follows: Where Mo and Vo are the desired mean and variance values, respectively. Normalization is a pixel-wise operation. It does not change the clarity of the ridge and valley structures. 4.2 FINGERPRINT SEGMENTATION In parallel with the normalization process fingerprint segmentation is also going on. Segmentation of fingerprint image is necessary so as to reduce the size of the input data, to eliminate undesired background , which is the noisy and to focus on area which is in favor of the central part of the fingerprint. The proposed fingerprint segmentation is further divided into three processes: skin color detection, adaptive thresholding, morphological processing. The algorithm proposed for skin color detection is: 1. The image is converted from RGB color space to YCbCr color space. Segmentation of the skin color regions becomes robust if only the chrominance component is used in analysis. Therefore, the variations of luminance component are eliminated by choosing the CbCr plane (chrominance components) of the YCbCr color space to determine the probable fingerprint region. The brightness component contained in the Y of the YCbCr will not be used. 27 2. Then, those images are low pass filtered to remove noise. The filter mask for the low pass filtering is as below and it can be said as “average filtering”. 3. Mean and covariance for Cb and Cr are determined by: 4. Region of interest is determined by using: After the region of interest is determined, the image is threshold into a binary image with adaptive thresholding. If the background has close resemblance to the skin color, it will be taken as fingerprint. To solve this problem, binary image processing technique is applied. The purpose of this process is to smooth, fill in, and/or remove objects in a binary image. Now, the binary mask from segmentation is multiplied with the normalized image. 4.3 FINGERPRINT EXTRACTION USING STFT ANALYSIS The fingerprint image may be thought of as a system of oriented texture with the local ridge orientation and ridge frequency varying slowly throughout the image. Due to this non-stationary nature of the image, traditional Fourier analysis is not adequate to analyze the image completely. So it is required to resolve the properties of the image both in space and also in frequency. The steps in this proposed analysis are: The image is first divided into overlapping windows during STFT analysis. The overlapping window is used to preserve the ridge continuity and removes ’block’ effects common with other block processing image operations. The image is assumed stationary within this small window and can be modeled approximately as a surface wave. Probabilistic estimates of the ridge frequency, ridge orientation and energy map are obtained after the Fourier spectrum of this small 28 region is analyzed The Fourier spectrum is represented in polar form as F(r,θ) A probability density function p(r,θ) and the marginal density functions p(θ), p(r) are defined as 4.3.1 RIDGE ORIENTATION IMAGE Assume that the orientation θ is a random variable that has the probability density function p(θ). The expected value of the orientation may then be obtained by performing a vector averaging. The resulted orientation image O(x,y) obtained from above equation is further smoothened using vectorial averaging and the smoothened image O’(x,y) is obtained. 4.3.2 RIDGE FREQUENCY IMAGE The average ridge frequency is estimated in a manner similar to the ridge orientation. Assume the ridge frequency to be a random variable with the probability density function p(r). The expected value of the ridge frequency is given by the following: 4.3.3 REGION MASK Simply, energy map is used as a region mask to get finer fingerprint segmentation. The region mask is obtained by using Otsu’s optimal thresholding technique. The energy map is obtained by: 29 The orientation image is then used in computing coherence image to prevent spurious artifacts caused by the discontinuities in the ridge flow at the block boundary especially at in the regions of high curvature close to the core and deltas which have more than one dominant orientation. 4.4 CORE POINT DETECTION To differentiate the entries of fingerprint images singular points, SPs are used. SPs are points that can be consistently detected in a fingerprint image and can be used as a registration point. Typically there are two types of singular points: core point and delta point. In this paper we only proposed the core point detection method. Fingerprint’s core point can be defined as the point of maximum curvature in the fingerprint image. 30 A fingerprint can have two structures, the global and the local structure. In the global structure the overall pattern of the ridges and valleys are considered where as in local structure the detailed pattern around a minutiae point is considered. A minutiae point is a position in the fingerprint where a ridge is suddenly broken or two ridges are merged. The global structure is used because it is more stable even when the fingerprint is of poor quality. Core points have special symmetry properties which make them easy to identify also by humans. Core Point detection can be done by using complex filtering [10]. The algorithm proposed for core point detection is: 1. Complex filter of order m are modeled by exp{imφ}. A polynomial approximation of these filters in Gaussian windows yield (x+iy)mg(x,y) where g is a Gaussian defined as g(x,y)=exp{-x2+y2 /2σ2}. 2. Now these filters are applied not directly to the original enhanced fingerprint image but they are applied to the complex valued orientation tensor field image Z(x,y)=(fx+ify)2 where fx is the derivative of the original image in the x-direction and fy is the derivative in the y-direction. 3. Filters of first order symmetry are used i.e. 4. The complex filter response is c=µexp{iα}, where µ is a measure of symmetry and is the “member” of that symmetry family. By using certainty measures µ1 and µ2 for core point and delta point respectively, we can identify an SP of type core if | µ 1 | > T1 and of type delta if | µ2 |> T2 , where T1 and T2 are threshold. 5. Then, multi-scale filtering is used to extract SPs more robustly and precisely compared to a representation at only one resolution level. The extraction of an SP starts at the lowest 31 resolution level and continues with refinement at higher resolutions. 6. The complex orientation field z(x, y) is represented by a four level gaussian pyramid. Level 3 has the lowest, and level 0 has the highest resolution. The core and the delta filtering are applied on each resolution. The complex filter response is called cnk where k=3, 2, 1 and 0 are the resolution levels, and n=1, 2 are the filter types. 7. In order to improve the selectivity of the filters, i.e. a filter should give a strong response only to one of the symmetries so we have used the following rules to sharpen the magnitude of the filter responses: With levels k=0, 1, 2, and 3. The complex filter response is cnk=snkexp{iαnk} where snk is a measure of certainty for that there is a symmetry of type n at resolution k, and αnk is how much the symmetric pattern is rotated to a fixed reference. After the core point is detected the enhanced image is cropped to size 200x200 in which the core point is at the center of the image. CHAPTER FIVE FEATURE EXTRACTION 32 5.1 MINUTIA EXTRACTION Most fingerprint identification methods use minutiae as the fingerprint features. The steps involved in minutiae extraction are smoothing, local ridge orientation estimation, ridge extraction, and thinning and minutiae detection. But for a small scale system, it is not efficient to process all the steps. So, we proposed Gabor filter-based feature extractor because its frequency and orientation representation are similar to those of human visual system. 5.2 GABOR FILTER Gabor filter helps in smoothing out noise and preserving true ridge valley structures. 33 f is the frequency of the sinusoidal plane wave, θk is the orientation of the Gabor Filter, σx and σy are the standard deviations of the Gaussian envelope along the x and y axes respectively. When core point is detected, the cropped normalize image is cropped again into size of 200 x 200 taking core point as center. Then images are sampled using Gabor filters and afterwards these filtered images are divided into set of 8 x 8 non-overlapping blocks. After that standard deviation is calculated which results in a scalar number and it forms Gabor feature of each image. All features collected are scaled to range of [0, 1] by where n: normalized feature. s: original feature. M: maximum value of all features. m: minimum value of all features. CHAPTER SIX FINGERPRINT VERIFICATION 34 6.1 CLASSIFIERS-BINARY Pattern recognition is the process of classifying input data into objects or classes based on key features. There are two classification methods in pattern recognition: supervised and unsupervised classification. Pattern recognition has applications in computer vision, radar processing, speech recognition, and text classification. Supervised Classification 35 The supervised classification of input data in the pattern recognition method uses supervised learning algorithms that create classifiers based on training data from different object classes. The classifier then accepts input data and assigns the appropriate object or class label. In computer vision, supervised pattern recognition techniques are used for optical character recognition (OCR), face detection, face recognition, object detection, and object classification. Unsupervised Classification The unsupervised classification method works by finding hidden structures in unlabeled data using segmentation or clustering techniques. Common unsupervised classification methods include: • K-means clustering • Gaussian mixture models • Hidden Markov models In image processing and computer vision, unsupervised pattern recognition techniques are used for object detection and image segmentation. Binary Classification Binary or binomial classification is the task of classifying the elements of a given set into two groups on the basis of classification Some typical binary classification tasks are: • medical testing to determine if a patient has certain disease or not – the classification property is the presence of the disease; • A "pass or fail" test method or quality control in factories; i.e. deciding if a specification has or has not been met: aGo/no go classification. • An item may have a Qualitative property; it does or does not have a specified characteristic 36 • Information retrieval, namely deciding whether a page or an article should be in the result set of a search or not – the classification property is the relevance of the article, or the usefulness to the user. 6.2 SVM- SUPPORT VECTOR MACHINE In machine learning, support vector machines (SVMs, also support vector networks) are supervised learning models with associated learning algorithms that analyze data and recognize patterns, used for classification and regression analysis. Given a set of training examples, each marked as belonging to one of two categories, an SVM training algorithm builds a model that assigns new examples into one category or the other, making it a nonprobabilistic binary linear classifier. An SVM model is a representation of the examples as points in space, mapped so that the examples of the separate categories are divided by a clear gap that is as wide as possible. New examples are then mapped into that same space and predicted to belong to a category based on which side of the gap they fall on. Fingerprint verification is done by SVM. SVM is a binary classifier which is based on the principle of structural risk minimization and maps an input sample to a high-dimensional feature space . SVM could optimally separate the two classes of genuine and imposters by constructing a hyper-plane. The training dataset is labeled as Where xi are Gabor features extracted belongs either to genuine or imposter class, yi is class label (+1 for genuine, -1 for imposter). Suppose we have a hyperplane that distinguish positive from negative examples. It is formulated as 37 where: w is normal to plane. b is the bias term. For finding hyper-plane which separate two classes which has maximum distance to closest point on each side of that plane, square of L2-norm of w where ||w||22 is imposed to inequalities(xi . w + b = 0) yi>=1 for all I is minimized. Each data point is transformed to a higher dimensional space for extending to non-linear boundaries. Proper transformation is required for achieving better separatibility between the two classes. For performing this transformation polynomial kernel function as formulated or radial basis function kernel as formulated in below are used. where n is order of polynomial. where σ is width of radial basis function. Another Method for verification is PCA (Principal Component Analysis) and Three Distance Measures [11]. Since the number of Gabor features extracted is huge, PCA is used for decreasing dimensionality of the feature vectors and retaining only those characteristics of feature vectors which contributing most its variance and eliminate later principal components. For matching three distance measures are used which are Manhattan distance, Euclidean distance and Cosine angle. The Manhattan distance (L1-norm) is addition of absolute difference between two feature vectors. It is formulated as follows: 38 The Euclidean distance (L2-norm) is the distance between feature vectors derived as a straight line and most commonly used calculation. The Cosine angle is cosine of angle between two feature vectors. It is calculated as: Let d is the distance computed from above 3 eq and threshold is set to T, matching is done if d < T and rejected when d ≥ T. CHAPTER SEVEN 39 MINUTIAE MATCH The minutia match determines whether the two minutia sets are from the same finger or not. It includes two consecutive stages: one is alignment stage and the second is match stage. 40 1. Alignment stage: Given two fingerprint images to be matched, choose any one minutia from each image; calculate the similarity of the two ridges associated with the two referenced minutia points. If the similarity is larger than a threshold, transform each set of minutia to a new coordination system whose origin is at the reference point and whose x-axis is coincident with the direction of the referenced point. 2. Match stage: After we get two set of transformed minutia points, we use the elastic match algorithm to count the matched minutia pairs by assuming two minutia having nearly the same position and direction are identical. 7.1 ALIGNMENT STAGE 1. The ridge associated with each minutia is represented as a series of x-coordinates (x1, x2…xn) of the points on the ridge. A point is sampled per ridge length L starting from the minutia point, where the L is the average inter-ridge length. And n is set to 10 unless the total ridge length is less than 10*L. So the similarity of correlating the two ridges is derived from: where (xi~xn) and (Xi~XN ) are the set of minutia for each fingerprint image respectively. And m is minimal one of the n and N value. If the similarity score is larger than 0.8, then go to step 2, otherwise continue to match the next pair of ridges. 2. For each fingerprint, translate and rotate all other minutia with respect to the reference minutia according to the following formula: where (x,y,θ) are the parameters of the reference minutia, and TM is 41 The following diagram illustrates the rotation of the coordinate system according to the reference minutia’s orientation: This method uses the rotation angle calculated earlier by tracing a short ridge start from the minutia with length D. And since the rotation angle is already calculated and saved along with the coordinates of each minutiae, then this saves some processing time. The following step is to transform each minutia according to its own reference minutia and then match them in a unified x-y coordinate. 7.2 MATCH STAGE The matching for the aligned minutia patterns needs to be adaptive since the strict match requires that all parameters (x, y, θ) are the same for two identical minutiae which is impossible to get when using biometric-based matching. 42 This is achieved by placing a bounding box around each template minutia. If the minutia to be matched is within the rectangle box and the direction difference between them is very small, then the two minutiae are regarded as a matched minutia pair. Each minutia in the template image either has no matched minutia or has only one corresponding minutia. The final match ratio for two fingerprints is the number of total matched pairs divided by the number of minutia of the template fingerprint. The score is 100*ratio and ranges from 0 to 100. If the score is larger than a pre-specified threshold (typically 80%), the two fingerprints are from the same finger. CHAPTER EIGHT 43 SYSTEM EVALUATION AND CONCLUSION 8.1 EVALUATION OF THE SYSTEM 44 As we can see in the graph shown below, when eliminating a step from the whole process or changing some of the parameters, the matching process is affected. Observations: 1. When altering in such an important step such as the image enhancement part, the performance quality of the system drops rapidly as the noise in the image is increased. Because when working with a biometric identification system, obtaining clear and noise free images is a really hard thing, so this step is usually needed. 2. For the binarization step, as explained earlier, using global thresholding may introduce a few problems and may lead to the elimination of significant details by mistake. Here, I tried using global thresholding, with 2 different thresholds, once using an intensity threshold of 120 and the second time using a value of 80. As we can see from the graph, setting the threshold at 120 (although it’s almost the average value for a gray-scale image) affected the system performance a lot and led to false non-match results, while setting a fixed threshold as low as 80 gave better results. Still, it remains better to use the adaptive threshold method because, although it consumes more processing time, it still guarantees the quality of the results. 45 8.2 CONCLUSION In this paper, a review of touch-less fingerprint recognition system, which can be an automated or a biometric digital camera based system is presented. We also presented number of comparisons between touch-less systems and the conventional fingerprint recognition systems. Further, this paper presented a modeled system that comprised of preprocessing, feature extraction and matching. Preprocessing is further subdivided into normalization, segmentation, enhancement and core point detection. The feature extraction might be based on minutiae extraction or image based method that is Gabor filter in which feature vectors are extracted. Moreover we have presented an effective verification technique that employs the SVM classifier and compares it with three distance measures. Future Scope We are looking forward for discovering a better enhancement method for fingerprint images captured by the digital camera by introducing some concepts of soft computing so that the output of preprocessing module increases the performance of system. We are also trying to discover more optimal methods for matching the fingerprint images. We are also directed towards removing the limitation due to digital camera and also to make this system to work with mobile phone so that fingerprint images can be captured by it 46
Purchase answer to see full attachment
User generated content is uploaded by users for the purposes of learning and should be used following Studypool's honor code & terms of service.

Explanation & Answer


Anonymous
Goes above and beyond expectations!

Studypool
4.7
Trustpilot
4.5
Sitejabber
4.4

Similar Content

Related Tags