Implementación de un algoritmo para el control de acceso basado en reconocimiento facial y de perfil.

Abstract

ABSTRACT: In the present degree project, the prototype of a biometric system based on facial recognition and the lateral profile of the face is presented. The recognition is carried out in real time from two angles in order to avoid possible violations. If it were only carried out based on the frontal image, it could be falsified with a photograph of the subjects authorized to enter. The system consists of an application that will run on a computer, two web cameras, an external light source, and the side panel that will have an appropriate background color for profile recognition to work. The algorithm obtains images of the user from the two cameras, automatically detects the face and profile using the Viola-Jones algorithm and extracts it for processing. For facial recognition, an MLP (Multi Layer Perceptron) multilayer perceptron-type neural network is used, and for lateral recognition, an algorithm is used that obtains the profile of the person who wishes to be authenticated. Users will first need to register in the program and then they can test the system authentication. In the testing phase, it was verified that the system achieves 90% effectiveness when detecting the user correctly.