Ph.D. Tezi Görüntüleme

      In this study, a new hierarchical age estimation system consisting of age group classification and age estimation modules is designed. To fit a common basis for the age groups used in the classification phase regardless of the methods and databases, and thus to facilitate the comparison of the studies with each other, the boundaries of age groups are designated according to the developmental psychologist Erik Eriksons “stages of psychosocial development” theory. With the usage of these age groups and fusing the features that are extracted with existing feature extraction methods in decision level both in classification and age estimation phases, a hierarchical age estimation system based on hybrid features was established. The classification accuracy and the age estimation performance of this system outperforms the previous works when using the FG-NET, MORPH-I and PAL aging databases which have very different characteristics.In the study, a model based age classification method is designed for the first time in the literature. For this purpose, age models representing the common characteristics of proposed age groups are produced using image morphing technique and classification is carried out with various distance metrics over these age models. Then classification accuracy is increased by combining the results of these metrics using a scoring method based on voting.