Conception et réalisation d’un système biométrique utilisant la reconnaissance faciale.

Abstract

Face recognition is an important function of surveillance systems to allow verification and identification of individuals of interest who appear in a scene captured by a distributed network of cameras. This project aims to use biometric images in order to improve authentication performance. In the first part, we propose a comparison of two methods of detecting the face in the image. In a second part we are interested in the identification process itself. A deep neural network approach is adopted. Our project was carried out thanks to numerous libraries among which tensorflow. Our work has also allowed us to solve the problem of real twins in facial recognition systems by proposing an on-board two-level security system.