Controle de acesso baseado em biometria facial

• Ano de defesa: 2010
• Autor(a) principal: Moraes, Jairo Lucas de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Espírito Santo; BR; Mestrado em Informática; Centro Tecnológico; UFES; Programa de Pós-Graduação em Informática
• Palavras-chave em Português: Biometria da face; Informação biométrica; Inteligência artificial; Computadores - Controle de acesso; Ciência da Computação; 004
• Link de acesso: http://repositorio.ufes.br/handle/10/4231

Face recognition is one of the most ordinary and natural task carried out by humans. Even though it is a simple task, it has proved to be a major challenge for artificial intelligence and computer vision researchers. Researches on object detection and recognition, and more specifically those related to human face, has greatly increased their demand in recent years mainly due to its employ in applications such as: public safety, access control, continuous authentication on computer networks, among others. This dissertation studies the feasibility of a system for access control using only facial biometrics as access key. In this case, access control would no longer be based "Something that the person has" or "something that the individual knows" but it becomes the person itself. To inquire the feasibility of an access control system based only in face biometric, we developed a prototype of this system that operates fully automatically, being able to detect a face in a static image or in a video and then perform the recognition of that face, with no human intervention. We use a well known approach (VIOLA, 2001) for the task of face detection and VG-RAM WNN (Virtual Generalizing Random Access Memory Weightless Neural Networks) for the recognition assignment. Lastly, we employed Bayesian probabilistic techniques for the access control problem. The obtained results are promising. The access control to given a resource was simulated for a number of 50, 100 and 200 users. For the set of 200 users the system was able to authenticate correctly 93.00% of the users with a FAR (False Acceptance Rate) of only 0.77%, for the set of 100 users the system was able to authenticate correctly 90.25% of the users with a FAR of 1.79%, and for the set of 50 users the system properly authenticated 93.11% of users with a FAR de 4.76%.