Reconhecimento de padrões faciais: uma síntese

Abstract

Facial recognition is a biometric task performed in a natural, instant and daily way by the human being, which recognizes people individually or in groups through visual contact. With the technological advancement, facial recognition techniques are present in a variety of environments, for example, in mobile applications, social networks, and in camera-monitored locations. Thus, the facial recognition task becomes an important and versatile resource in the market, being used to promote both the entertainment to the users as well as the security and monitoring of the individuals who attend a certain place. Despite the technological advancement and rise of facial recognizers, the methodology for creating and configuring such a system is complex, requiring the correct development of tasks that are chained in its results, thus becoming the final recognizing face approach. The current literature on facial recognition has few works focused on the complete process for the generation of this type of system, with most of them concentrating on tasks or specific stages of the whole process. The aim of this work is the elaboration of a synthesis about the creation of a facial recognition system, presenting all the necessary steps for this and their respective techniques, conceptual approaches and related nuances. The facial recognition algorithms Eigenfaces, Fisherfaces and Local Binary Patterns Histograms (LBPH) will be addressed, listing the activities of face detection; pre-processing of facial images; feature extraction; and final face classification. Finally, the results of several experiments with the mentioned algorithms will be presented, analyzing the performance of the tasks of facial recognition and the quality in the individuals final identification.