الفهرس | Only 14 pages are availabe for public view |
Abstract Age-Invariant Face Recognition (AIFR) is one of the most crucial computer vision problems, e.g., in passport verification, surveillance systems, and missing individuals, so person identification despite age progression is a challenging issue, in this thesis we present two age-invariant face recognition systems. The first proposed system includes four stages: Preprocessing, feature extraction using transfer deep learning, feature fusion using Multi-Discriminate Correlation Analysis (MDCA), and classification. The main contribution of the first system (MDCA-AIFR) is the utilization of MDCA fusion which significantly reduces the feature space dimensions and improves the recognition rate. The second proposed system is based on the optimization of deep learning features using a Genetic Algorithm (GA) procedure, without preprocessing phases. The GA-base optimization is designed in order to select the most relevant features to the problem of identifying a person based on his\her different age images. The second system (GA-AIFR) shows an ability to improve the recognition rate over the first system (MDCA-AIFR), without the need of computationally expensive preprocessing registration phases. Our experiments in both systems are performed on two standard face-aging datasets, namely, FGNET and MORPH datasets, respectively. Both Age-invariant face recognition systems achieve high Rank-1 recognition rates. In addition, the experiments show their privilege over other state-of-the-art techniques on same data. These results show the promise of the proposed systems for personal identification despite aging process. |