الفهرس 
INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 107 |  |  |
| | | from | 107 | | |
Abstract
Security will always remain an issue of a great importance in business and in personal life. Biometric-based security systems have several advantages over traditional authentication systems, and it is gaining higher interest each day from researchers and market. Most of the security systems, as the biometric system, take its decision based on the question of (who are you?) and this is done by using the unique characteristic for each person which differ from any other person. Biometrics is described as the science of recognizing an individual based on his/her physical or behavioral traits. The human iris is a physiological biometric characteristic. It is claimed to be one of the best biometrics as it has unique features and it is complex enough to be used as a biometric signature. No two irises can be identical in two different persons even if they are twins, and an iris never changes during a person’s life time. The aim of this thesis is to study and compare some different algorithms of iris localization and segmentation for iris images as well as feature extraction and recognition of different persons based on their iris images. Studying and analyzing the performance of proposed different algorithms enabled us to enhance the recognition rates of iris recognition systems. In this work, we were trying to reach the best recognition rate in iris recognition system through the optimization of the neural network weights and biases using two different optimization techniques particle swarm optimization and gravitational search algorithm. First, the iris image was preprocessed and enhanced using histogram equalization. Second, the iris image was segmented using circular Hough transform which is more accurate compared to other segmentation techniques. Second, normalization using Daugman‘s rubber sheet model was performed to convert from the Cartesian coordinates to polar coordinates. Next, feature extraction using Haar wavelet transform which was proved that it is more efficient than Gabor wavelet transform. The feature vector was further reduced in dimension using principle component analysis to reduce the convergence time of the neural network. Finally, the obtained feature vector was fed into the feed forward neural network optimized by two different methods of optimization which are the particle swarm optimization and the gravitational search algorithm trying to get the optimal weights and biases and hence higher recognition rates. The effectiveness of these two optimization techniques in enhancing the recognition rate was compared. This thesis is organized as follows; Chapter 1 presents an introduction to biometrics as well as the neural network and optimization. Also, an overview to the thesis is introduced. Chapter 2 give a literature review on some of the iris recognition systems. Chapter 3 discusses the first three stages of the proposed iris recognition system which are the (segmentation), (normalization) and (feature extraction and reduction). Chapter 4 highlights on the last stage which is the classification, this chapter focuses on feed- forward neural network and two optimization techniques (particle swarm optimization and gravitational search algorithm) and applying these two methods of optimization on the neural network. Chapter 5 discusses the implementation of different iris recognition stages (Segmentation, Normalization, Feature extraction and classification) in iris recognition technology as well as analysis of the results of comparisons between different methods of optimization on the neural network. Chapter 6 represents the conclusion and the recommendations for future work.