Utilização e Comparação dos Métodos Eigenfaces, Fisherfaces e LBPH no Processo de Reconhecimento Facial

Abstract

Digital security has become an increasingly relevant topic in light of technological advancements and the rise of cyber threats. In this context, facial recognition emerges as a promising solution for authentication and access control. This work presents a comparison between the Eigenfaces, Fisherfaces, and Local Binary Patterns Histograms (LBPH) methods for facial recognition. The study is justified by the growing demand for accurate and efficient facial identification methods in contexts requiring high levels of access control and information protection. Initially, the theoretical foundations related to computer vision, face detection, and facial recognition are explored. In the methodological process, the algorithms are implemented using the OpenCV library in Python. The image acquisition phase is carried out using two datasets: one created by the author and another obtained from the internet. The image preprocessing stage includes conversion to grayscale, resizing, and the application of filters. This step is crucial, as it provides the necessary foundation for the subsequent stages of face detection and recognition. The algorithms are evaluated in terms of accuracy, precision, recall, and F1-score. The results indicate that the LBPH algorithm demonstrates the best overall performance, with an accuracy of 98%, followed by Fisherfaces with 92%, and Eigenfaces with 88%. Additionally, the main advantages and limitations of each algorithm are discussed. This study contributes to a comparative understanding of facial recognition methods and provides a valuable reference for selecting the most suitable algorithm for access control systems in indoor environments.