Fingerprint recognition using Fuzzy inferencing techniques

Roshini Velamuri Guide: Sarkodie-Gyan Thompson A report submitted as a partial submission of the requirements of Fuzzy logic and Engineering University of Texas El Paso April 20, 2006

Abstract: The paper develops the fingerprint recognition system using fuzzy inferencing techniques. The two principle components in this recognition system are the administrator, where templates are stored and the fingerprint recognizer. We do not use the intermittent steps of Image processing techniques like enhancement the image and reduction of the noise. In comparison to other recognition systems designed, our approach has an advantage that is the proposed system is cost effective. I. Introduction:
In 1956, researchers discovered [2] the uniqueness of human fingerprint: there are no two fingerprints that are alike. Soon after this discovery, many major law enforcement departments have adopted the idea of recording the fingerprints of the criminals so that their records are available and later they could determine the identity of the criminals from the left over fingerprint smudges at the crime scene. Despite the inventive methods to improve the efficiency of manual recognition system, it did not meet the demands imposed by increasing work loads. Furthermore, the overall cost involved was high and this lead to initiation of research for automatic recognition systems. The recent advancements in bio-informatics and imaging techniques have led to the development of Biometric Systems. A Biometric System is a system capable of capturing a biometric sample (Identifier); extract the data; compare the data with reference templates and give a decision indicating whether the identity claimed by the individual is matched. The different identifiers that have been used for designing various recognition systems are fingers (esp. index and thumb image), hands, feet, faces (facial images), eyes (iris image), ears, teeth, veins, voices, signatures (style), typing, styles, gaits and odors [2]. These systems solve the identification problem. Example: Forensic DNA has used the iris recognition technology that could differentiate between individuals of identical DNA such as monozygotic twins. Depending on the requirements of the application, a match between the biometric identifiers and characteristics of the application are made. The biometric identifiers narrowed for most of the civilian applications is the fingerprint as the fingerprint analysis system is proven (shown in table 1) to be cost effective, robust, and efficient.

Distinctiveness

DNA Ear Face Facial thermogram Fingerprint

H M H H M

H M L H H

H H M L H

L M H H M

H M L M H

Performance

Universality

Permanence

Biometric Identifier

L H H H M

L M H L M

circumvention

Collectability

Acceptability

Gait Hand geometry Hand Vein Iris Keystroke Odor Retina Signature Voice

M M M H L H H L M

L M M H L H H L L

L M M H L H M L L

H M M M M L L H M

L M M H L L H L L

H M M L M M L H H

M M L L M L L H H

Hence, a lot of research has been initiated in this developing the fingerprint recognition system. The Federal Bureau of Investigations technique of verifying and identifying a fingerprint from set of minutiae is widely adopted in the existing fingerprint recognition systems. However, these systems have their draw backs. Most of these systems require some of the preprocessing steps such as transforming to binary image, thinning etc. This makes the task of matching the fingerprint with the data base a tedious job. In this project, we use direct gray scale minutiae detection by a back up procedure to eliminate the false minutiae. For matching, we use fuzzy evolutionary programming. . The advantage of fuzzy
logic over conventional method is its ability to represent and process human knowledge in form of fuzzy ifthen rules.

II. Description of our approach:

Creating the template and storing them in the data base: The data base stores the templates of fingerprint. Step 1: Perform quality test. The steps involved are. o Capturing image o Performing quality test. Step 2: Minutiae extraction, Each minutia is represented by location (x, y) and local ridge direction . The process of minutiae detection begins with finding a summit point on the ridge, then this point continues tracing until a minutiae determined by ending, bifurcation is encountered. A ridge and its representation is shown in below figure1

Figure 1 To find a summit point. Start at a point X = (x1, x2) Compute the direction angle (We used gradient method to determine the ridge orientation)

Tracing the ridge: Start from summit point Move pixels along to another point y Find the next summit point Z that is the local maximum of orthogonal to set x= z Until the termination point is reached The false minutiae or termination is determined by using fuzzy if then rules. Some of the rules are: If the ridge end is close and opposite direction then delete both of them If the bifurcation point is end of one branch then delete both of them If the termination is close to more than two other terminations then delete all of them Once the minutiae is determined, we store (as shown in figure 2) them as a template in the data base

Figure 2 Fingerprint Matching: Several approaches to match the fingerprint are roughly classified into three families. Correlation- based matching , In this method two fingerprints are superimposed and correlation between pixels is computed for different alignments Minutiae- based matching, most popular and widely used method. In this method the matching essentially consists of finding the alignment between template and the input minutiae sets that results in the maximum number pairings.

Ridge feature- based matching, in this method a comparison in term of the features are extracted from ridge pattern.

Our approach of Fingerprint matching In our approach, we calculate the matching degree of the target fingerprint with the images in a database and decide if it belongs to particular individual. The method of matching degree is described below, Consider two fingerprints represented by P = { Where

P ,K P
1

}Q={

q ,K q
1

for 1 <= i<= m, 1<=j<=n The key task is to find the transform F = (s, , x, y) which transforms the minutiae set P into set Q. S represents the scaling factor and represents the scaling factor and (x, y) represents the translation into xy- plane The new transform is defined by , of minutiae , defined by

To find the degree of matching we associate F (p) into fuzzy set. The member ship function defined as Where And

represents a subset defined by

We allow a degree of tolerance in matching F(p) and q. the tolerance is low when the minutiae are apart The matching degree between F(p) and Q is defined by

The computations are made until it is stabilized. There are two steps involved in determining the match between P and Q, which are 1. 2. Estimation of rotation Fuzzy evolutionary programming

Estimation of rotation is described as below. Divide the interval [pi, -pi] into K subintervals Set an integer array c[1 K] and a real array [1 K] and initialize them to zero For i = 1m and each j= 1.n Find k such that Increase c[k] by 1 And increase [k] by Find c[k*] such that c[k*] is maximum Let defined by

Fuzzy evolutionary programming Generate a population of m minutiae

Where parameters are randomly chosen

For each K = 1...m generate

Where

Where the normal distribution With variance Moreover, is inversely proportional to to ensure we are in the permitted intervals

Compute

until the entire process is stabilized

The Equipment used is: o o o A CCD camera to capture the images. Work station to accommodate the software / database required (finger print images) to be used in the application. A Scanner to get the fingerprint images,

III. Applications:
1. 2. The application of this system can be adopted in wide range such as banking, e commerce and access control. The other application areas include immigration check points, national ID, voter and drivers license, registration and time and dependence.

IV. Milestones:
The following are the projected activities during first half period of the project. Q1- week 1., week 2 - gather information about the various possible methods in fuzzy evolutionary programming Q2- week 3 - Design the algorithm for minutiae detection.

The following are the anticipated tasks to be completed for the next half period of the project. Q1- week 4., week 5., week 6 Purchase and install new workstation Complete the minutiae matching algorithm and create data base for fingerprint Extend this algorithm to read different fingerprint images and create templates.

Q2- week 7., week 8. Test the algorithm for various fingerprint images. Extend the scope of the new design approach to other civilian applications

VI. Results: The algorithm designed for minutiae extraction is run for two iterations and the results obtained are

V. Conclusion: The recognition system uses gray scale ridge tracing. This is backed up with a validation algorithm developed using fuzzy inferencing techniques. The results are expected to be highly effective with relatively clean fingerprints. To handle images with noise we are working to add some more pre processing based evolutionary programming. Also attempt to connect this system directly to live fingerprint scanner. VI. Reference: [1] Tu Van Le; Ka Yeung Cheung; Minh Ha Nguyen; A fingerprint recognizer using fuzzy evolutionary programming System Sciences, 2001. Proceedings of the 34th Annual Hawaii International Conference on Jan 3-6 2001 Page7 pp. [2] Patricia Melin, Angeles Quezada, and Oscar Castillo Pattern Recognition using the Fuzzy Sugeno
Integral for Response Integration in Modular Neural Networks Dept. of Computer science, Tijuana Institute of technology Tijuana Mexico Neural Networks Ensembles IEEE Transactions on pattern Analysis and Machine Intelligence 1990

[3]. D Maltoni, D. M.A.K. Jain, Hand book on finger print recognition system Springer science, ISBN: 0-387-954317-7, 2003. [4] Salil Prabhakar, Fingerprint classification and matching using a Filter bank, Ph.D. Dissertation, Michigan state University, 2001.

[5]. Emiroglu, I.; Akhan, M.B.; Pre-processing of fingerprint images Security and Detection, 1997. ECOS 97, European Conference on 28-30 April 1997 Page147 151. [6] Lin Hong; Jian, A.; Pankanti, S.; Bolle, R.; Fingerprint enhancement Applications of Computer Vision, 1996. WACV '96. Proceedings 3rd IEEE Workshop on 2-4 Dec. 1996 Page(s):202 - 207 Digital Object Identifier 10.1109/ACV.1996.572056. [7] Lin Hong; Yifei Wan; Jain, A.; Fingerprint image enhancement: algorithm and performance evaluation Pattern Analysis and Machine Intelligence, IEEE Transactions on Volume 20, Issue 8, Aug. 1998 Page(s):777 - 789 Digital Object Identifier 10.1109/34.709565. [8]. Simon-Zorita, D.; Ortega-Garcia, J.; Cruz-Llanas, S.; Gonzalez-Rodriguez, J.; Minutiae extraction scheme for fingerprint recognition systems Image Processing, 2001. Proceedings. 2001 International Conference on Volume 3, 7-10 Oct. 2001 Page254 - 257 vol.3 Digital Object Identifier 10.1109/ICIP.2001.958099. [8]. Tu Van Le; Ka Yeung Cheung; Minh Ha Nguyen; A fingerprint recognizer using fuzzy evolutionary programming System Sciences, 2001. Proceedings of the 34th Annual Hawaii International Conference on Jan 3-6 2001 Page7 pp. [9]. Jain, A.K.; Uludag, U.; Hiding biometric data Pattern Analysis and Machine Intelligence, IEEE Transactions on Volume 25, Issue 11, Nov. 2003 Page(s):1494 - 1498 Digital Object Identifier 10.1109/TPAMI.2003.1240122.