A Review on Face Recognition-based Access Control Systems

This paper presents a review on the state-of-the-art of the access control systems based on face recognition. The review reveals the following: i) More than fifty percent of the related contributions have been published in the last five years. ii) The most used techniques to achieve the face recognition are neural networks, principal component analysis, local binary pattern, and linear discriminant analysis. These techniques have been applied mostly to improve the performance of the recognition accuracy or recognition rate and less for addressing variations in illumination, face spoofing, information security, privacy, face occlusion, computational time, classification performance, small sample size, and recognition with low-resolution images, pose variations, and expression changes. iii) Other several techniques, including Viola-Jones, hidden Markov model, and Gaussian mixture model, have been less used to deal with the aforementioned problems (except recognition with low-resolution images, pose variations, and expression changes) and recognition with retouched or rotated images. iv) New challenges in the face recognition-based control systems appeared due to the occlusion of the faces with masks by COVID-19. Also, open challenges and future work where artificial intelligence could be harnessing are given.