International Journal of Computer Engineering and Applications, Volume VI, Issue I April.14

ISSN 2321-3469

2D Fingerprint Matching using Multi- dimensional Artificial Neural Networks

Ms. Vishakha R. Bhadane, Dr.M.V.Deshpande,Mr. Prashant Mishra

Department of Computer Engineering,MPSTME, SVKM’s NMIMS, Shirpur, Maharashtra, India

Abstract- Biometrics is a technology which identifies a person based on physiological or behavioral characteristics. Fingerprint identification and recognition is one of the biometric methods available that has been widely used in variousapplications due to its reliability and accuracy in the process of identification and authentication of a person. The main objective of this work is to develop fingerprint identification and recognition system. The system consists of three main parts; are image acquisition, image processing and "identification". The fingerprint images captured sensor format and stored in the database in the process of image acquisition. This image is the commitment to share 500 DPI. These images are then amplified in stage image processing by performing gray level enhancement, Gabor filtering, image sharpening, edge detection, segmentation and thinning processes, then the minutia extraction method using the cross number and alignment images. We present a hybrid algorithm that uses both minutiae matching (point) information and texture (region), adjustment information of fingerprints. Once the image processing could then be inserted into the back propagation neural network as input to train the network. After training, the neural network is ready to perform the identification and recognition function (matchingprocess). A neural network has been successfully developed to detect and recognize the essential part of fingerprint images.

Keywords-Fingerprint Recognition, back propagation,multi-dimensional ANN

  1. INTRODUCTION

Fingerprints are the most used technique biometrics for personal identification. There are two main applications include fingerprints: verification of fingerprints and fingerprint identification [1]. While the purpose of the verification of fingerprints to verify a person's identity, the purpose of fingerprint identification is to identify an individual. In the last three decades, automatic verification of fingerprints is becoming more widely than other techniques of biometrics, such as face recognition and signature recognition.

Biometrics is the emerging field in technology with unique and measurable "physical and behavioral” characteristics that can be processed electronically by creating recognition , run control and automatic face recognition . These physical properties include the appearance of the face, fingerprints, hand geometry , handwriting , iris , retina , voice and veins . Fingerprints have been used for over a century and is the most widely used form of biometric identification . Fingerprint identification is mainly used in the field of artificial intelligence. It is popular because of its easy access , low price of fingerprint sensors , non -intrusive scanning, and relatively good performance. In recent years , there have been significant improvements in performance were achieved in commercial automatic fingerprint identification . The footprint of an individual is unique and remains unchanged over a lifetime. No two people have the same set of fingerprints. This property makes an excellent fingerprint biometric identifier. So it is one of the popular and effective means to identify an individual and used as forensic evidence . A fingerprint is formed from an impression of the pattern of ridges on a finger. A ridge is defined as a single curved portion and a valley is the region between two adjacent ridges .Usually , there are two prominent types of minutiae (ridge ending and ridge bifurcation ) , forming a fingerprint . The minutiae are local discontinuities in the flow pattern ridge provides the features used for the recognition . The details such as the type , orientation and location of the minutiae taken into account when performing minutiae extraction [1 ] . Here , we propose a fingerprint matching system based on artificial neural network (ANN ) . The rest of the paper is organized as follows : Section II provides a brief description of a general fingerprint recognition system . Section III provides the background principles associated with the operation of the proposed model . All the experimental results and related discussion is provided in Section IV - V. This paper concluded by summing up the work in Chapter VI.Some of the relevant literatures cited from [21] [18].

  1. ARCHITECTURE OF FINGERPRINT MATCHING SYSETM

Fingerprint Identification System based can operate either identification or verification operation. Fingerprint identification refers to one-to - many matches where fingerprint input image of a person associated with any standards that exist in the database. Confirms the identity of a person. Fingerprint Identification System follows four basic stepsis Image Acquisition /Registration, image enhancement preprocessing, feature extraction and Matching.

Fig. 1. General block diagram of Fingerprint recognitionSystem

Figure understands the automated fingerprint authentication comprising the following phases [5]:
a) Fingerprint Acquisition / Register
b) Fingerprint Image Enhancement
c) Extraction of Minutiae
d) Minutiae Matching
e) fingerprint classification / Authentication

A. Image Acquisitions

At this stage, the image of the fingerprint was arrested for the first time with the help of sensors. Acquired images can be blurred or may contain noise, which lowers image quality directly affects the rate of return of fingerprint recognition system. The fingerprint image acquired can vary the position, direction and stretching degree [5].

B. Pre-processing

After capturing an image from the image or image enhancement preprocessing applied to the image. Sometimes it can contain image noise while recording process; noise can be removed using filters in treatment / improvement phase. Images must be normalized; this can also be done in the pretreatment step. You must have completed this phase.

C. FeatureExtraction

After the process of feature extraction, preprocessing is performed. The phase characteristics of image export function exported as Ridges, valleys and minutiae, singular points (loops, core, coils and delta), etc. These features are useful for identification or verification of individuals. The features extracted from the captured images stored in the database for further process of matching.

D.Matching
Next step is the connection process after feature extraction. Phase matching identifies similarities between the mainpattern recognition and pattern. Different approaches pairing. The input images provided in the system associated with the primary standards that exist in the database . Matching is completely dependent on whether the systemperforms identification or verification. While recognizing that performs one - used to-many matching approach ,where the fingerprints of a person fits all the available templates under various one-to - one fight is to control where the input image of a person associated only with a template person who claims to be.

III FINGERPRINT CLASSIFICATION

Fingerprint classification defines the global standard representations of fingerprints. Worldwide representations include positions of the points (e.g. , the nucleus , and d) a fingerprint. A standardclassification of fingerprints are categorized in the following six categories: whorl, right loop, left loop,arch , twin loop, arch and scenes . Also contains one or more regions where the lines consume different ridge shapes (curvature, termination, etc.). These areas (called singular anomalies or areas) can be classified into three typologies: loop, delta, and whorl[7]

Fig. 2. Structure of fingerprint. (a) Ridges and valleys on afingerprint image; b) singular regions (white boxes) and corepoints (small circles) in fingerprint images [4]

  1. SURVEY OF FINGERPRINT MATCHING TECHNIQUES USING ARTIFICIAL NEURAL NETWORKS

Different classification techniques have been proposed by various authors in the field of Artificial Neural Networks.ANN used to give greater accuracy and learning pace. It is basically used for classification purposes in image processing. Fingerprints are traditionally classified into categories based on the information in global standards ridges. The classification of fingerprint into groups reduces the need to match an input fingerprint with the entire fingerprint database during the process of detection and recognition, and therefore to reduce computational requirements. Two classifiers,ie , the mean and 3 K - nearest neighbor , used to extract the classifier has four different types of fingerprints, ie Arch, Left Loop, Loop right, or spiral . The method achievehigh classification accuracy and quick to produce results . Tan, X. ,Bhanu, B. , Lin, Y, used a typical algorithm - learning using genetic programming (GP) to learn and capabilities have evolved image processing operators to classify fingerprints. The primitive operators used were simple and easy to calculate . These bodies were divided into vectors and calculating feature vectors generation. This classification method may be found to be effective with the quality of fingerprint images [10] .

Maheswari and Chandra [22], shows the classification scheme of fingerprints operation Fuzzy artificial neural network. The characteristics of the fingerprint, such as singular points, the positions and orientations of the core and delta points from a binary fingerprint image captured by the sensors. The method used to produce the good results of classification using fuzzy neural networks. An algorithm used for two machine learning algorithms presented by [15]. They used Support Vector Machine (SVM) and Recursive Neural Networks (RNNs) during classification. The RNNs are trained on a structured representation of the fingerprint images and also used to extract a set of distributed features incorporated in SVMs. SVMs are combined with error correction coding scheme , which exploits the information contained in ambiguous fingerprint images .

Kant and Nath presented an approach that improves the speed, efficiency of fingerprint matching algorithm during the same record. For this reason , the hard points of fingerprints, as delta and core classifiers used and grouped into any of the other six categories. [20] According to the Wei L, anomaly detection can be used to increase the accuracy of the classification algorithms and proposed a method for searching mannerisms use index delta field Poincare. Used these abnormalities , a quick rule -based classification algorithm was proposed to classify fingerprints into five categories, arch, tents , left loop, right loop , whorl and double loop . The detection algorithm searches the direction field which has the larger direction changes to get the defects [16]. Wei, Yonghui and Fang [18] proposed an approach based structure based on characteristics curve ridgelines, used for classification of fingerprint with other fingerprint images available in the database. The algorithm mainly uses the direction for the classification of fingerprint ridgelines. In this method, the classifier is first calculated the total change of direction ridgelines is grouped according to their shape. Grouped ridgelines along with the extracted singular points used for classification of fingerprint arch, tents, left loop, right loop , whorl and double loop courses . Combining singular points and orientation image information for fingerprint classification proposed by [19]. Algorithm says those singular points and limit non -linear orientation and the final feature vector consisting of the coefficients of the model orientation and uniqueness information. This led to a compact feature vector used as input to a Support Vector Machine (SVM) to perform image classification.Chandra Maheswari and has used low dimensional features obtained from the feedback -based line detector to classify fingerprints into five classes ( arch , left loop , right loop , whorl , arch and scenes ) . The detector line was a cooperative dynamic system that gives oriented lines and maintain multiple directions at points where different oriented lines meet . The feature extraction based on characterizing the distribution of orientations around the fingerprint. Three types of classifiers used namely , support vector machines , nearest neighbor , and neural network [14]. An algorithm used for two machine learning algorithms presented by [5]. They used Support Vector Machine (SVM) and Recursive Neural Networks (RNNs) during classification . The RNNs are trained on a structured representation of the fingerprint images and also used to extract a set of distributed features incorporated in SVMs. SVMs are combined with error correction coding scheme , which exploits the information contained in ambiguous fingerprint images .

  1. PROPOSED WORK

The above discussed methods have some limitations which have solved by many researchers, but there are still some other issues in fingerprint recognition and matching systems like computational time and efficiency of recognizing the faces from large databases.

Figure 3: Overall system description diagram

Figure understands the automated fingerprint authentication comprising the following Modules:

1] Data Capturing Module

2] Image Enhancement Module

3] Minutia Extraction

4] Recognition and Matching Using MDANN

5] Decision Making Module

1. Data Capturing:

Data logging is offenly, as it is called, also, obtaining fingerprints. Altitude fingerprints can be either offline (inked) or Online (Live scan). In a method inked fingerprint of a finger taken initially impregnated in a paper, which is then scanned. This method usually produces images of very poor quality due to the non- uniform dispersion of ink and therefore not practiced online touch [11]. In this online fingerprinting system sensor is a capacitive or optical fingerprint scanner which is more expensive, patented fingerprint sensor with USB Optical resolution 500DPI. The image size is captured form the fingerprint reader is 260x300 pixels. The most important part of the system is the living image of the fingerprint capture section which is also called as smart capture that capture high quality fingerprints from dry , wet , scarred and age is difficult to detect fingers more accurately and reliably. After taking Pictures of the fingerprint reader will be given to that section image enhancement.

2. Fingerprint Image Enhancement / Preprocessing

A critical step in automatic fingerprint matching is automatically and reliably extracts minutiae from the input fingerprint images. However, the performance of a minutiae extraction relies heavily on the quality of input fingerprint images. To ensure that the performance of an automatic fingerprint matching system is strong in relation to the quality of the fingerprint images, it is essential to incorporate a fingerprint enhancement

A. Segmentation

Segmentation is the process of telling the background area of the foreground regions. It is observed that the background areas are low gray scale variation compared to the foreground region. Thus, a method based on the threshold variation is used for the partition [24].

V(k) = (1)

where V(k) isthevarianceand M(k) isthemeangray-levelvalue of theblock k. I(i,j) isthepixelvalueatthepixel(i,j).

B.Normalization

Normalization is done in order to standardize the dynamic variation in the gray-level values.

N(I,j) = if I(I,j)>M(2)

Otherwise(3)

where ‘N(i,j)’ is the normalized gray-level value and I(i,j) is the gray-level value at pixel(i,j). ‘M’ and ‘V ‘are the estimated mean and variance of I(i,j) respectively and ‘’ and ‘’ are the desired mean and variance values respectively.Basically the best normalization technique now a day's is mostly used is histogram equalization. Histogram Normalization is one of the most commonly used methods for preprocessing. In image processing, the idea of ​​equalizing a histogram is to stretch and redistribute the original histogram using the entire range of discrete levels of the image, in a way that an enhancement of image contrast is achieved. The most commonly used histogram normalization technique is histogram equalization where one attempts to change the image histogram into a histogram that is constant for all brightness values. This would correspond to a brightness distribution where all values ​​are equally probable. For image I (x, y) with discrete k gray values ​​histogram is defined by ie the probability of occurrence of the gray level i is given by:

P (i) = (4)

Where i∈ 0, 1…k −1 grey level and N is total number of pixels in the image. Transformation to a new intensity value is defined by:

= = (5)

Output values are from domain of [0, 1].To obtain pixel values in to original domain, it must be rescaled by the K−1 value.

C.Filtration

Along with the field orientation, the ridge frequency is also required for the construction of the filter Gabor. This represents the frequency of the local ridge of the fingerprint. Gabor filters optimally capture both local orientation and frequency information from the image of the fingerprint. By tuning a Gabor filter with a certain frequency and direction, can be obtained by the local frequency information and orientation [7] [8]. It is suitable for extracting information from images texture [9].

Another symmetric Gabor filter has the following general form in the spatial domain [10]:

G(x, y; θ, f) = exp (6)

x’ = x sinθ+ y cosθ (7)

y’ = x cosθ- y sin θ(8)

wheref is the frequency of the sinusoidal plane wave along the direction θ from the x-axis, and x δ and y δ are the space constants of the Gaussian envelope along x and y axes, respectively.

D.Binarization

Binarization is to convert the grayscale image to a binary image that is either zero or oneasthepixelintensity value using a certain threshold.

E.Thinning

This is the final stage of image enhancement. This is a process which erodes morphological image successively until a skeletal image of the fingerprint which is one pixel wide isobtained. This image is in turn used for the extraction process of the minutiae [24].

3 Minutia Extractions

The enhanced image is used to extract the minutiae points of the fingerprint. There are a number of features that could be used for identification of fingerprints, but mostly minutiae points is limited to two types namely the ending ridge and the ridge bifurcation. Ridge endings is the point where the curve ends ridge and ridge bifurcations are the points where the curve ridge splits from a single path to two in a Y - junction. The minutia points extracted from the enhanced fingerprint image by using the concept of Crossing Number (CN). The value CN for a ridge pixel P is given by the equation [24].

CN = (9)