Controle de Acesso com Reconhecimento Facial: Uma Revisão Sistemática da Literatura

With the aim of identifying how facial recognition is used in monitoringand attendance control systems, this study presents a systematicliterature review (SLR) on the topic. It was found that themost employed Artificial Intelligence technique was ConvolutionalNeural Networks (CNN). Regarding algorithmic approaches, theHaar Cascade Classifier was used in most studies. Concerning thedatasets, proprietary databases were the most frequent. In terms ofthe number of images in the datasets, the analyzed studies rangedfrom small experimental sets to databases with millions of images,and the employed hardware ranged from affordable devices to highperformanceGPUs. The test scenarios explored both controlledenvironments, with variations in lighting and angles, and practicalapplications in real-world settings. Finally, accuracy rates rangedfrom 78.40% to 100%, with an average of 96.02%, while precisionrates varied between 80% and 100%, with an average of 94.98%.The main contributions of this study lie in identifying the mosteffective techniques, the challenges faced, and the gaps that mayguide future studies.