Fingerprint Feature Extraction Using Convolution and Particle Swarm Optimization Algorithms

Most of the existing fingerprint extraction systems are based on the global features and detailed characteristics of fingerprints, which have a weak performance in cases of poor quality fingerprint images, such as the fingerprint image is incomplete. In order to improve recognition accuracy, reliability and quickness to identify the fingerprints a new trend has been opened by using swarm intelligence techniques in biometric field. Therefore, particle swarm optimization techniques (PSO) are used in this paper to build fingerprints authentication system. A fast fingerprint identification method based on the convolution transformation and Particle Swarm Optimization algorithms proposed. The convolution algorithm was used to extract the convolved feature and then found the optimal solution from this feature by using Particle Swarm Optimization algorithm. Experimental results show that, the proposed method has a high efficiency in extracting features from fingerprints, strong strength, and good accuracy for recognition.


Introduction
With the increased use of computer technology for data processing and access to sensitive or personal data also becomes more important.The use of PIN (personal identification number) is not more reliable to secure these types of data; therefore, the use of biometrics is the perfect solution.Biometric technologies can block unauthorized access or misuse of any system (computer networks, cellular phones, ATMs, PCs, Smart cards, etc.).Personal identification numbers and passwords can be forgotten, and determine identification symbolism can be stolen.Therefore, identification of biometric systems has more interest [1].
Matching fingerprints is a standout amongst the most huge and reliable approaches to identify the person.There are two major applications on a comparison of fingerprints: fingerprint identification and verification of fingerprints.While the motivation behind the fingerprint identification is to determine a person's identity, the objective of the fingerprint verification is to install the identity of the person.As a rule, fingerprintercognition includes comparing the select fingerprint with a huge number of fingerprints put away in a database.And that takes a tall time.To minimize the search time and low computational intricacy, and is often used rating of fingerprints to divide the database into smaller subsets [2].Yager [3] a recent review of the methods of rating of fingerprints.Usually, it is classified fingerprints into five noteworthy classes, known as "Henry classes", specifically Arch, Tented Arch, Whorl, Left Loop and Right Loop.Maltoni [4] Non-direct bends coming from skin flexibility, sensor clutter and Low quality of fingerprint image make the problem of fingerprint rating very difficult so that a many frameworks have been proposed to manage this problem.The primary element that characterizes the "Henry classes" is the ridge flow pattern design, which on a basic level can be described by the number and sorts of singularities in the direction field (i.e., ridge direction field).Kawagoe and Tojo [5] delta and Core focuses are the fundamental elements utilized as a part of guideline based methodologies.However, these systems are subject to failure when his characters do not show or cannot be extracted.Wang [6] has presented one of these ways to take advantage of the orientation information.Chong [7] for the same purpose ridges represented by splined curves were employed.Cappelli [8] an auxiliary methodology utilizing apportioning of the orientation field into homogeneous areas.Zhi-Hui Zhan [9] an adaptive particle swarm optimization (APSO) presented an algorithm which has the best optimal efficiency in search of classical particle swarm optimization (PSO).Gabriel [10] two improvement procedures are assessed for this assignment: genetic algorithms (GA) and particle swarm optimization (PSO) utilizing standard GA and make a distinction PSO.In this paper we present a proposed mechanism for fingerprint feature extraction using a new way where research include two parts: The first part is the use of convolution transformation and in the second part we used the way of Particle Swarm Optimization algorithm to extract the best feature from fingerprint image.

Convolution
The mathematical concepts of convolution and the kernel matrix are used to apply filters to images, to perform functions such as extracting edges, lines, corners and reducing unwanted noise.Convolution filtering is used to modify the spatial frequency characteristics of an image.unique positions from top to bottom that the mask can take.Corresponding to these positions, each feature in the output will contain 3x3 (i.e., (N-k+1) x (N-k+1)) elements.The convolved value obtained by summing the similar elements between the mask and the input image [11].

Particle Swarm Optimization
Interest was growing in researches of particle swarm optimization (PSO) which is proposed by theEberhart and Kennedy in 1995 [12].In comparison the PSO with many other optimization algorithms, PSO has an advantages of strong global search, faster convergence rate, few adjustable parameters, simplicity of the algorithm, and ease of implementation [13].Due to its many advantages, the particle swarm optimization algorithm can be used widely in the fields such as function optimization, the model classification, machine study, neural network training, the signal processing, vague system control, and automatic adaptation control [14].

The Proposed System
The block diagram of the proposed system is shown in Figure ( 1)

B. Image Binarization
Binarization can be defined is that a process of digitizing the images by converting a gray level image to a binary image.Therefore, binarization converts the image from a 256-level image to a 2-level image that gives us the same information.Normally, the value '1' is given to an object pixel, while a background pixel is given a value of '0'.At the end, a binary image is created by coloring each pixel white or black, depending on a pixel's label (black for 0, white for 1) [17].

C. Transformation using Convolution
Fingerprint as the natural images have property stationary, meaning that the statistics of one part of the image are the same as any other part.This suggests that the features that we learn at one part of the image can also be applied to other parts of the image, and use the same features at all locations, take the learned features and convolve them with the larger image, thus obtaining a different feature activation value at each location in the image.As in the algorithm 1.

D. Feature Extraction using PSO
The feature is extracted from the image of a fingerprint using particle swarm algorithm (PSO) which is a mathematical way to improve the problems that in the pictures by trying to replicate a solution to improve the candidate with respect to certain the masterminding of quality .And called the name of the particles as they move around the space of the image to find the best neighbor based on a simple calculation formulas are affected quickly and the position of the particle.It is expected that the squadron is moving toward a better solution, thereby feature extracting of a fingerprint.The goal of this algorithm is to get the optimal solution and the result is better -across simulate the behavior of birds in search of better food and so any system that relies on this algorithm will be formed at the beginning of the random grouping of random solutions.After recording information , fingerprints capture and the features extract of the fingerprint using an PSO algorithm ,feature exraction using PSO explains in algorithm 2.

Experimental Results
In this section will display the results that were obtained from the experiments after the implementation of the convolution transformation and PSO algorithm on a number of fingerprints images.

Different Mask Effect
The feature extraction is the important step in proposed method, these feature should be unique for each voter.transformation using convolution is method used here and it depend on mask such: 3*3 convolute in widow such 5*5, these will produce 10 different cases these will count its histogram and get max one as shown in table1.The first mask is [1 1 1 ; 0 0 0 ; 1 1 1].As explain in previous table 1 in spite of the case number 6 is it in most image but the value of histogram are differ as shown figure 2. The second mask used is [1 1 1 ; 1 0 0 ; 0 1 0] used and represent the other 10 cases of features explain in table 2. The previous table 2 show the majority max histogram is in case number 5also different value as shown in figure 3. The third mask used is [1 0 0;0 1 0;0 0 1] used and represent the other 10 cases of features explain in table 3.  The previous table 3 shows the max histogram is in case number 3 also different values as shown in figure 4.

Vol: 13
No:4 , October 2017 DOI : http://dx.doi.org/10.24237/djps.1304.276CP-ISSN: 2222-8373 E-ISSN: 2518-9255 Convolution of an input with one mask (kernel) produces one output feature, and with H masks independently produces H features.Starting from top-left corner of the input image, each mask is moved from left to right, one element at a time.Once the top-right corner is reached, the mask is moved one element in a downward direction, and again the mask is moved from left to right, one element at a time.This process is repeated until the mask reaches the bottom-right corner.For the case when N = 5 and k = 3 , there are 3 unique positions from left to right and 3 PSO is a population based search procedure in which the individuals, called particles, adjust their position to search through the search space.Each particle Pi has a position vector ( , ... )and a velocity vector ( , ... ).For each iteration the particles learn from its own previous best position and the best position of all the other particles in the swarm, and updates it's' velocity and position.The update equation at (k + 1) -iteration can be written as follows: …(1) …….(2) Where c1 and c2 are the cognitive and social scaling parameters respectively, and r1 and r2 are random numbers in the range of [0, 1].Vol: 13 No:4 , October 2017 DOI : http://dx.doi.org/10.24237/djps.1304.276CP-ISSN: 2222-8373 E-ISSN: 2518-9255 As there are few parameters to adjust for the Particle Swarm Optimization (PSO) and its updating procedure has only simple arithmetic operations, it is a good choice for fast optimization.The Particle Swarm Optimization (PSO) algorithm provides a better control of the adaptation process.It allows the user to control the number of particles, particles' maximum velocity, and ranges of each coefficients, tolerance level, and stability[15].

Figure 1 :•
Figure 1: The block diagram of applied fingerprint feature extraction algorithm

5 Fingerprint
G E 1 I M A G E 2 I M A G E 3 I M A G E 4 I M A G E