Facial recognition and landmark localization in niqab-wearing individuals pose challenges for computer vision systems due to occlusions. Traditional facial recognition systems rely heavily on visible facial features for accurate identification. When these features are partially or fully obscured, as with niqabs, models struggle to extract key landmarks, resulting in reduced accuracy. This is largely because most deep learning models are trained on non-occluded faces, making it difficult for them to generalize to occluded scenarios. Recent approaches, such as aware face alignment algorithms, ensemble regression trees, and CNN-based features, have improved accuracy under these conditions. The review also highlights the effectiveness of multi-task learning, attention mechanisms, and generative adversarial networks (GANs) in enhancing detection accuracy. Additionally, it explores applications in security, healthcare, and human-computer interaction, as well as recent advances in kinship verification, feature fusion methods, and the impact of ageing on facial verification. The ongoing need for continuous improvement in handling occlusions, particularly with niqab-wearing individuals, is emphasized to develop more reliable facial recognition systems.