Thesis (MSc.) - Cyprus International University. Institute of Graduate Studies and Research Department of Electrical and Electronic Engineering

Includes bibliography (sheets 50-52)

ABSTRACT In today's digital world, face recognition can have many uses in a variety of fields and subjects. This detection can speed up and simplify the process in many applications, as well as significantly improve the security of sensitive locations. For this reason, several studies have been conducted in this field and various methods have been proposed. By examining these studies, it can be concluded that there is still no reliable method that can perform automatic face recognition with high reliability, and shortcomings are still observed. One of the factors that has reduced the accuracy of the results of the previously proposed methods is the high volume of data. Therefore, by reducing the volume of this data, the computational load of the algorithms can be increased due to the elimination of less valuable data and accuracy in detection. For this purpose, the researchers have been able to solve the feature selection problem, which is an np-hard problem, using innovative and meta-heuristic algorithms. But these methods still have drawbacks that make the feature selection one of the most complex issues in face recognition. Due to the importance of this issue in this dissertation, we examine the application of particle swarm algorithm (PSO) to optimally select a feature in the face recognition problem in which the extracted features for the face recognition system are calculated using the SIFT descriptor. The SIFT descriptor will help to extract and correctly identify face images and assign them to a person in the relevant database. Therefore, the purpose of selecting a feature is to choose the best feature from the SIFT descriptor to extract the feature. The results obtained from this method indicate its usefulness, and its high accuracy shows that this method can be used practically in different places.