Iris segmentation using illuminance texture-based features

Dissertation for Master Degree (MSC)

Electronics Orientation

Dissertation abstract (including summary, objectives, implementation methods and obtained results):

A biometric system based on the unique characteristics of each person automatically identifies people. A system includes several steps to identify a person. These steps are 1) taking a picture 2) separating and Isolation of the iris region in the captured image 3) Normalization 4) Feature extraction and matching. In this project, we are trying to talk about the separation of the iris area from the captured image. The separation step is very important and sensitive because the codes based on these areas are not accurate subsequently and this wrong information is given to the system at the registration stage. The research presented in this thesis is to present two solutions for accurate iris zoning based on the characteristics of the lighting texture. The first solution is using the active contour method. In this method, using the values ??obtained from the input image, it gives us the equation of an ellipse, which represents the desired iris region. This program was tested on 100 images from the collection of known reference images (Kasia) and the success rate was about 89% with an error The second solution is to use wavelet and genetics. By using the wavelet coefficients of the input image, we get a circle that is placed on the smallest possible area by the genetic program, which is the iris area. The success rate is about 83.5% with an error of 16.5%. It has been shown that identifying people through iris images with the proposed methods has acceptable results. Attention and accuracy are among the available methods. Introduction Problem statement: A biometric system automatically identifies people based on the unique characteristics of each person. Identification through iris images is considered as one of the most reliable and reliable methods. Most of the commercial products in the field of iris are made based on the registered algorithm proposed by Dogman, which have 100% identification capability, but the algorithms presented in the articles are tested under favorable conditions and without considering practical problems. The research presented in this thesis includes the zoning of iris images using features based on lighting texture. To check the functionality of this algorithm, a set of Cassia images was used as a test. The presented method includes iris image segmentation based on two active contour methods and genetic algorithm, which can determine the inner and outer borders of the iris along with its center and radii. The results obtained by using several types of proposed algorithms have finally been tested on CASIA images, which include 756 images of 108 people, and the success rate is about 89% with an error rate of 11%.

1-2       Research objectives:

Here, our goal is to perform better zoning using the features of the lighting texture

or in other words, to improve the zoning with the help of methods.

1-3 The importance of the research topic and the motivation for choosing it:

Identification through iris images is currently considered as one of the most reliable and capable methods. This thesis includes iris segmentation using features based on lighting texture. 1-4 Main questions and hypotheses of the problem: Research question: Is it possible to design and implement a new algorithm that can be identified through the iris of the eye with high reliability? First, the border of iris and sclera should be determined, then finding the pupil with the help of wavelet transformation and finally testing the techniques on the database, these parameters have a great impact on the identity recognition performance. It seems that if these parameters can be accurately obtained during the zoning, the zoning can be done with the least error. As a commercial contract has been found in China belonging to 1200 years ago, and the end of the contract was signed by the parties.Still in the Far East, the fingerprint of the owner of the book can be seen on the back of the books where the name is written. For the first time in 1870 AD, the use of measuring different parts of the bones of prisoners' bodies was introduced by a French person named Burliton, and this system was used in the United States of America until 1920. In 1880 AD, the use of fingerprints and faces was suggested. In the identification card that was issued to people in Germany after World War II, the owner's fingerprint was also recorded. With the advancement of signal processing science in the 1960s, in addition to the previous ones, voice and signature were also used. Retinal capillaries were the next thing that was implemented in the 1980s. Although the use of iris was proposed in 1936, its use became practical in 1993. In today's modern society, due to the rapid growth of the economy, there is an increased need to recognize or confirm the identity of people in investment. When granting access is necessary to do something, for example entering a forbidden place or the presence and absence of people in a company, granting permission is generally in the form of issuing privileges to a specific person or group. In this chapter, we will review biometric technology. Today, with the development of technology and the need for more security in societies, it has made people look for faster and more reliable ways to identify people. Identification methods have become more accurate and reliable over time. Identification methods are usually of three types:

1- A person's assets, such as keys, magnetic cards, and smart cards.

2- Through knowledge, such as passwords and identification numbers.

3- Biometric properties.

Biometrics using automatic or semi-automatic use of physiological or behavioral characteristics of the body. Man and its analysis is to identify the individual. Each of the used methods has weaknesses and strengths that can be eliminated by combining them with other security methods. The biometric system is designed in such a way that instead of using something you have like a key or something you know like a password, it uses what the human being has made. Things that are never lost, stolen or forgotten. Also, in the identification process, a person must be present and it is not necessary to memorize or remember information. For this reason, experts consider this method of identification to be much safer and more reliable than any other method. In the biometric system, we are faced with two principles: [1]

Identity recognition

Identity verification

The biometric system is able to confirm whether a person is the person he claims to be or not. Also, by receiving the person's information and comparing it with the files saved by the deceased person, the identity can be recognized. In fact, confirmation is a one-to-one comparison. For example, in hand verification, the scanned hand pattern is compared with the hand of the person who claims it. While identity recognition requires that a person's characteristic is compared to the same characteristic of all the people who are stored. For example, it compares your fingerprint with all stored fingerprints, and finally, if it is already stored, it will recognize the person. The difference between these two is well shown in (Figure 2-1) and (Figure 2-2). Identity.

According to the technology used in them, existing old methods can be divided based on 3 main properties [1] Proprietary-Knowledge-Biometric.

Regarding the problems of old methods, for example, bank ATM systems, when a person wants to use his card, he needs to enter his personal identification number in order to withdraw his money. that the person has with him, which is a potential problem, for example, the card in this system can be stolen, the password is vulnerable, especially if this person behind you is looking at your hand while entering the password. However, it will be difficult for someone who has stolen the card to use the card without knowing the password.