Extract Unique Personal ID Using Fingerprint

Security documents like certificates, land revenue documents, etc., have only the individual’s name, address, and in some cases a photo as means of personal identification. This makes criminal impersonation an easy task. Therefore, many of these limitations can be eliminated by incorporation of better methods that can be used to verify identity by measuring and analyzing human characteristics instead of things the individual may have (smart card) or know (password). The main goal of this paper is using the fingerprint technology to generate the unique ID based on the core point of the fingerprint of an individual that can be used for identifying person identity. The finger-print’s minutia features are extracted with the core point as the reference based on the seven moment invariants, Then the extracted features are converted into numerical value, This numerical value is used as the unique ID for printing in the security documents for the personal identification. If the fingerprint image does not contain the core point such as the plain arch and tented arch patterns then we must create the center point in this fingerprint image and this is considered as the core point. The minutia features will be computed with this point as the reference point and this is used to generate a unique ID.


Introduction
Highly important documents such as certifications, passports, driving licenses, and land revenue documents have a degree of security that makes forging rather a difficult task.
However, once a malicious person who has good technical experience gets the possession of such a document, they can simply perform a criminal impersonation.This is very simple due to the fact that documents like these have only name, place of residence, and a picture of the owner for the ID.Every single one of these details is under the risk of duplication.The usage of biometric characteristics will be helpful in reducing the risk of duplication.Nowadays, the usage of biometrical features in these documents is quite uncommon.However, in order to stop impersonation and avoid the forgery of this kind of documents, the correct solution will Biometrical techniques in general, take a specific portion of the human's body to distinguish an individual and those techniques are very precise.The person's biometrical features that are deduced have to be somehow transferred to the documents [1].
Biometrics means measuring of individual physiological or behavioral features of an individual for the sake of identifying that individual.Biometrics is an individual feature which is a part of the human being.Therefore, there's no necessity to worry about remembering any passwords, or carrying any form of document for identification.Biometric features can be divided in two basic kinds.Physiological character: Which is concerned with the shape of the body and therefore it is different from one person to another.Examples of that are fingerprints, face recognitions, hand geometry and iris recognitions.Behavioral character: This is concerned with the individual's behavior such as signatures, key stroke dynamics and voice.Behavioral features might change as the person ages [2].
Personal identification is mainly split into three kinds depending on what the person possesses (such as a credit card or keys), by something they know (such as a password or a PIN code) or by physiological or behavioral features.The third method is called biometrics and the six most widely known characteristics consist of face, voice, iris, signature, hand geometry and fingerprint.It has been proven and is widely known, the fact that everybody has an individual fingerprint that doesn't vary with the change of time.Every individual's finger is of its own unrepeated pattern; therefore, any finger might be used in identifying an individual successfully [3].
Fingerprints are of the most predominant biometrical features.Fingerprint is the representation of the friction ridges on finger.A friction ridge is an elevated part of the epidermis on the surface of the finger and they are permanent along the life.Two fingerprints, even in the case where they're taken from identical twins, could never be identically similar.Some of the most common biometrical characteristics such as Fingerprints, Iris and face recognition and fingerprint are the most popular feature.It's used in individual distinguishing in many personal and civilian implementations due to its characteristics such as uniqueness manner.The ridges in arch manner drawn with no backward turn or re-curve.The patterns of loops are the ones where the ridges make a backward turn with not wits.Therefore, the center of the print seems to look like a hairpin in loop pattern.The ridges in whorl pattern form a highly complicated pattern that has two or more deltas.The composite pattern includes a mix of two or more arches, whorl and loop patterns [1]. .

Figure (1): Basic patterns of finger-print ridges
In the pattern of a fingerprint there are specific individual points, such as core and delta that are mainly marked as the singular points.Singular points are the ones where there is a sudden variation in the ridge patterns.In finger-prints, Minutiae, the breakings in the patterns of ridges that interrupt the usually smooth flow of ridges, are important characteristics that are utilized in order to contrast one print from others.There are a number of kinds of Minutiae such as ridge endings, ridge bifurcations, short ridges, islands, spurs, bridges, and so on [1].This research reports a simpler mechanism for developing a unique ID through the extraction of the minutia features depending on the fingerprint's core point of a person for printing in the documents for the individual identification.
The paper is arranged in the following order: Section 2 explains concepts of moment invariants.Section 3 illustration of the proposed system presented.Section 4 gives the results acquired from the evaluation of simulation and performance seen in section 3. Lastly, section 5 concludes the analysis.properties which uniquely identify its form.Invariant recognizing of shape is done by classifying in the multi-dimensional moment invariant feature space.A number of methods for the derivation of invariant characteristics from moments for recognizing objects and for representation were developed.These methods are identified by their definition of moment, like the kind of the data that is implemented and the technique for the derivation of invariant values from the image moments.Hu was the first one to determine the algebraic base for the 2D moment invariants and implemented their applications in recognizing shapes.They have been initially implemented in Aircraft shape recognition and have proved to be fast and dependable (Dudani, Breeding and McGhee, 1977).These values of moment invariants don't change with respect to translating, scaling and rotating of the object [5].
Representation of region moment interprets a normalized grey level image function as a probability density of a two dimensional random variable.Characteristics of this random variable may be represented with the use of statistical characteristic-moments [Papoulis, 1991].Supposing that a pixel with values greater than zero represents areas, moments can be used in black and white or grey level area description.A moment that has the order (p + q) depends on scaling, translating, rotating, and even on grey level transformations and is depicted by [6]:
Translation invariance can be achieved if we use the central moments where the   values can be calculated from equation (7) .

The Proposed System stages
The proposed system contains two stages: feature extraction and unique ID generation.They are described in the following sections.

Feature Extraction
Feature extraction is defined as the process of converting a captured biometric sample, i.e. fingerprint, in to a unique, distinctive and compact form so that it can be compared to a reference template.Feature extraction is a significant step to improve the efficiency of personal identification.This stage consists of two phases training phase and testing phase.
Each phase has specific steps.Where total RW is the total ridge width of the fingerprint image, max W is the maximum value of the frequency of occurrence of the white ridge width, max B is the maximum value of the frequency of occurrence of the black ridge width and K is the integer value identifies how many lines that the fingerprint's feature has.Depending on trial-and-error, best integer value for K is 6.If K is more or less than 6 then the result is incorrect.Step12: Find the largest match and mark it as the fingerprint feature point on fingerprint image.End

Unique ID Generation
After finding the core point and the minutia points there will be a big number of minutiae.
However, just a little number of minutiae that is completely individual for that specific fingerprint is necessary in generating the unique ID.Therefore only the minutiae that lie within a Region of Interest (ROI) distributed in the area that surrounds the core point are taken.Here, the ROI is a square with an appropriate pixel area that is centered on the core point.There are two cases in the fingerprint image; in the first case the fingerprint image contains the core point such as loop and whorl, while in the second case the fingerprint image does not contain the core point like arch.Step8: Calculate the distance between each point and the rest of the other feature points which marked in the fingerprint image by using the Euclidean distance equation ( 14).Step9: Choose the less distance between two minutia points and find the midpoint between them by using the midpoint equation ( 15) and this is considered as the core point.Step10: Repeat steps 2 to 6 in order to generate a unique ID.End

Experimental Results of the proposed system:
This section illustrates the phases of implementing the proposed system:

Feature Extraction
In the training phase, samples of fingerprint's features were chosen.These samples consist of 40 sample images, 10 sample images for each fingerprint's features (core, delta, ridge bifurcation and ridge ending).The system will train all the samples images and store the values in the database.This phase runs only one time at the beginning of running the program.
After these sample images of each fingerprint's feature are converted in to binarized images, now the seven invariant moment processes are applied to these binarized images as described   In the testing phase, the system will identify an individual person by testing his fingerprint image.This phase is done every time we want to identify a fingerprint image.In this phase the binarization process will be applied to the original fingerprint image that was described in section (3.1.2.a).applying parametric equation of a circle as shown in Figure (5).Table (8) shows the results of applying the unique ID generation process.While in the case of the fingerprint image does not contain the core point as described in section (3. 2. b), the unique ID generation process must calculate the distance between each point and the rest of the other minutia points which are marked in the fingerprint image by using the Euclidean distance equation.After finding the less distance between any two minutia points and computing the midpoint which is considered as the core point of the fingerprint image, now the region of interest in fingerprint image and draw circles in clockwise direction in region of interest by applying parametric equation of a circle will be applied as shown in Figure (6).Table (9) shows the results of applying the unique ID generation process. .
of Moment Invariants Moment invariants are common characteristics used in the processing of images, remote sensing, recognizing shapes and classifying.Moments are capable of providing object images Where  ,   are the co-ordinates of the region's center of gravity (centroid), which can beobtained using the following relationships: of the binary,  means the region area (see equations (1) and (2)).Scaleinvariant features can also be found in scaled central moments   (scale change ̷ = ⍺,  ̷ =⍺ ) )  A group of 7 invariant moments are possible to be deduced from the second moment and the third moment which were introduced by Hu.As the formulas that are listed below, Hu deduced the expressions from algebraic invariants applied to the moment producing function under a rotating operation[7].

3. 1 . 3 . 1 . 2
1Training Phase: The training phase contains three steps.a. Sample Image Binarazation: Ten samples for each fingerprint's feature (core, delta, ridge bifurcation and ridge ending) that act as training database are converted in to binary form as pre-processing step in order to be used in the seven invariant moments in subsequent steps.Vol: 13 No:3 , July 2017 DOI : http://dx.doi.org/10.24237/djps.1303.271BP-ISSN: 2222-8373 E-ISSN: 2518-9255 b.Computing seven invariant moments of fingerprint's features samples: compute the seven invariant moments of the binarized sample images of each fingerprint's feature which were produced by the binarazation process in the previous step.c.Computing variance of fingerprint's features samples: The variance of each sample can be achieved by finding the maximum and minimum values of each moment of the samples and compute the mean of each moment of the samples by applying this equation: Mean = (max + min)/2.Then compute the variance of each sample by applying this equation   = ∑ ( () − () ).  = Testing Phase: The testing phase contains seven steps.a. Fingerprint Image Binarazation: compute the binarazation of the original fingerprint image which is obtained from testing database as pre-processing step.b.Region of Interest (ROI) Extraction: Crop the binarized of the fingerprint image into an ROI in order to separate foreground region and background region.c.Computing Total Ridge Width: The total ridge width of the fingerprint image can be computed by applying this equation: Total RW = (max W + max B)*K.
d. Fingerprint Image Partition:After computing the total ridge width, we will divide the fingerprint image (ROI) into blocks with equal size of w x w and detect blocks size based on the total ridge width that is computed in the previous step.The fingerprint image partition can be performed pixel by pixel.
No:3 , July 2017 DOI : http://dx.doi.org/10.24237/djps.1303.271BP-ISSN: 2222-8373 E-ISSN: 2518-9255 e. Computing seven moment invariants of blocks: compute the seven invariant moments of the blocks of the fingerprint image that are obtained from applying the fingerprint image partition process.f.Computing variances of blocks: The variance of each block can be calculated from applying this equation:   = ∑ ( () −   () ).  = where   is the variance of each block and   is the seven means of the fingerprint features stored in database g.Variance Matching: After calculating the variances of each fingerprint's feature stored in database and the variances of each block which were produced previously.The variance matching process first compares and matches the variances of each block with the variances of each fingerprint's feature stored in database.Next find the largest matches, these largest matches are considered as the fingerprint feature points and mark them on the fingerprint image.Algorithm (1): Fingerprint Image Features Extraction Input: a number of samples of each fingerprint feature, Fingerprint image.Output: the largest matches as the fingerprint feature points and mark it in the image.Begin Step1: Initialize fingerprint feature samples in a binary form.Step2: Compute the seven invariant moments as mentioned in equation (8) of each fingerprint feature sample.Step3: Find the maximum (Max) and minimum (Min) of each moment of the fingerprint feature samples and compute the mean of each moment of the fingerprint feature samples by applying equation (9) Step4: Compute the variances of each fingerprint feature sample by applying the equation (10).Step5: store the mean and variances of each fingerprint feature sample in the database for later use in the testing phase.Step6: Initialize fingerprint image in a binary form.Step7: Crop the binarized of the fingerprint image into an ROI in order to separate foreground region and background region.Step8: Compute the white ridge width and the black ridge width of the fingerprint image and find the histogram of the white ridge width and the black ridge width and then compute the total ridge width by applying the equation (11) Step9: Divide the ROIs into blocks with equal size of w x w and detect blocks size based on the total ridge width which computed in step 8. Step10: Compute the seven moments as mentioned in equation (8) of each block and compute the variance B of (12) Vol: 13 No:3 , July 2017 DOI : http://dx.doi.org/10.24237/djps.1303.271BP-ISSN: 2222-8373 E-ISSN: 2518-9255 each block by applying the equation (12).Step11: Compare and match of the   of each block with the   of samples stored in database.

a.For
The fingerprint image contains the core point: In this case the unique ID generation algorithm must apply the Region Of Interest (ROI) centered around the core point so that only the minutia points (ridge bifurcation and ridge ending points) which are inside ROI are taken, The size of the ROI can be determined by finding the distance between the core point and the rest of the other minutia points (ridge bifurcation and ridge ending points) by using the Euclidean distance equation and selecting the required number of the minutia points that have the less distance between them and the core point so that only these minutia points are inside ROI.The required number of the minutia points can be determined by the system user.By applying parametric equation of a circle, a circle can be defined as the locus of all points that satisfy the equations[8]: x =   + r cos Ɵ y =   + r sine Ɵ If we find minutia points located within the boundaries of these circles then we find their own angles.This process continues until the end of all the minutia points contained within the Region Of Interest (ROI).Each angle of the minutia points is converted to fixed length which consists of three digits.Depending on the priority of the finding the minutia points on the circles, all the angles are connected with each other to generate the numerical values.These numerical values are considered as a unique ID b.The fingerprint image does not contain the core point: In this case the unique ID generation algorithm must calculate the distance between each point and the rest of the other minutia points which are marked in the fingerprint image by using the Euclidean distance equation.In the Euclidean distance, if   (  ,  ) and   (  ,  ) then the distance is given by[9]:D (  ,   ) = √(  −   )  + (  −   ) Choose the less distance between two minutia points and find the midpoint between them by using the midpoint equation, if   (  ,  ) and   (  ,  ) then the midpoint is given by [10]: =(  +   )/ =(  +   )/ The Midpoint is P(x, y) = ( (  +   )/,(  +   )/) This midpoint is considered as the core point of the fingerprint image.Then repeat the same steps that have been applied in the case of the fingerprint image contain the core point in order to generate a unique ID. number of feature points that marked in the fingerprint image, Output: A unique ID.Begin Step1: If the fingerprint image contains the core point Step2: Apply Region Of Interest (ROI) that resides around the core-point so that only the minutia points which are inside ROI are taken.Step3: Draw circles in clockwise direction by applying the equation (13) Step4: If find minutia point lie within the boundaries of the circle then find the angle of this minutia point.Step5: All angles which found in step 4 is converted to fixed length consist of three digits.Step6: Depending on the priority of the finding of the minutia points on the circles, all the angles are connected with each other to generate the numerical values.These numerical values are considered as a unique ID.Step7: If the fingerprint image does not contain the core point.
in section (3.1.1.b).After computing the seven invariant moment values for each sample of the fingerprint's features, now we will compute the maximum value, minimum value and mean value for each moment of the samples and compute the variance of each sample as previously explained in section (3.1.1.c).Tables (1), (2), (3) and (4) show the results of applying the seven invariant moments, max, min , mean and variance processes to the binarized sample images of each fingerprint's feature (core, delta, ridge bifurcation and ridge ending).Vol: 13 No:3 , July 2017 DOI : http://dx.doi.org/10.24237/djps.

Figure ( 3 )
Figure (3): Fingerprint image partition pixel by pixel After calculating the variances of each fingerprint's feature stored in database and the variances of each block which were produced in the previously step, now will Perform variance matching as described in section (3.1.2.g). Figure (4) illustrates the results of extracting features from fingerprint image.Table (7) shows largest match for feature points in this table, feature type is the last column where 1 for core point, 2 for delta point, 3 for bifurcation and 4 for endpoint.Vol: 13 No:3 , July 2017 DOI : http://dx.doi.org/10.24237/djps.1303.271BP-ISSN: 2222-8373 E-ISSN: 2518-9255

Figure ( 4 )
Figure (4): Features extraction results in fingerprint image.Red box denotes the core point, blue triangle denotes the delta point, blue circle denotes the end point and green circle denotes the bifurcation point