Analysis of Face Recognition Technology From 2D to 3D in the Field of Security

This review discusses the development of face recognition technology in the field of security, focusing on deep learning and Two dimensions (2D), Three dimensions (3D) models and their multi-modal fusion. Among them, the recognition accuracy of the deep learning model taking Convolutional Neural Network (CNN) model as an example for static image and dynamic video can reach 91.7% and 86.7% respectively. Research on the currently widely used 2D and 3D models shows that for 2D and 3D multi-modal mixed models, the verification accuracy of natural faces can be as high as 99.74% under a strict error acceptance rate of 0.001. This technique can be applied to high-security environments such as airports, and the recognition rate based on Regularized Kernel Canonical Correlation Analysis (RKCCA) can reach 85.19%, which is 15% higher than linear Canonical Correlation Analysis (CCA). In order to solve the problems in 3D face reconstruction, a Goldstein branch method was introduced, which achieved 97.10% accuracy. Exceeds traditional techniques and reduces the risk of deception. However, although the technology has made great breakthroughs, it is still faced with large dependence on data, high computational complexity, and difficulties in the deployment of edge devices. This review highlights the development potential of face recognition in the field of security, while advocating for peaceful solutions that can reconcile accuracy, efficiency, and ethical issues.