Research on Face Anti-Spoofing Algorithm Based on 3D-DQLBP

With the increasing popularity of face recognition technology, face-spoofing frequently occurs, resulting in security risks and economic losses. The research on face anti-spoofing methods is particularly important. For video attacks, a face anti-spoofing method based on 3D-DQLBP is proposed. By introducing the time axis as the factor to further refine the DQLBP, a 3D model of human face video id created, the feature extraction of a human face video is improved to three-dimensional. Finally, use support vector machine (SVM) for classification. On the basis of not adding DQLBP dimension, the fusion algorithm and the feature point extraction method are improved to effectively speed up the operation speed. The experimental results show that the algorithm obtains the detection results with HTER of 0, EER of 0.14 and AUC of 99.85 on the Replay-Attack database. Compared with the face anti-spoofing frontier algorithm, it is analyzed that the algorithm has a development prospect.