Thesis (MSc) - Cyprus International University. Institute of Graduate Studies and Research Information Technology Department

Includes bibliography (sheets 56-63)

ABSTRACT
Face biometrics have been recognized as a feature for identification for many years,
and most researchers and scholars recognize these biometrics as one of the most
reliable and accurate biometrics. But in recent years, these biometrics have been used
to identify other human characteristics such as age, race, gender, and so on. Regarding
the vast range of applications for facial recognition in identification, significant
research has been done in this field, although it falls short of ideals. With the advent
of COVID19 at the end of the twentieth century, the challenge of masking face
recognition was created. According to extensive research on face research in this field,
existing face-based identification methods have received less attention in various mask
challenges. The primary goal of this research is to find answers to the following
questions of masked face recognition systems. Therefore, it is necessary to invent a
method that has a good response to stable changes and time, in order to recognize the
masked faces. This study uses a "Real World Masked Face Dataset" image database.
The method presented in this research, after preprocessing, Morphological Operations
has been utilized to derive features from pictures of people's faces by using the KNN
descriptors. To classify the properties extracted by the descriptors, two families of
super plane and neighborhood classifiers k are the nearest neighbors.
Finally, after implementing the ideas considered in the employed method and
comparing the results of the classification, 72.5% for the desired database for
classification with k nearest neighbor have been obtained for classification.
Keywords: Masked Face Recognition, Face Biometric