Study On Face Anti-spoofing Detection

The facial biometric technology of low-tech devices such as video(replay attack)is vulnerable to spoofing attacks,and facial anti-spoofing is essential to prevent the facial recognition system from being compromised.After the above analysis,first of all,this thesis introduces the methods based on manual design on the face anti-spoofing detection in recent years,the methods based on deep learning,and the methods based on mixed features are discussed and analyzed.Then,this thesis also proposes a method of multi-feature fusion,using convolutional neural networks to learn multiple features from different cues in face images for anti-spoofing detection,and the depth map can spatially distinguish the depth information between true and false faces;The optical flow map can distinguish the dynamic information between real and fake faces in time;the residual noise map distinguishes the difference between the noise components of the primary imaging of the real face and the secondary imaging of the fake face.The fusion of the three features not only uses space,Time multi-dimensional clues make up for the shortcomings of a single clue,and at the same time improve the generalization ability of the model.Compared with the existing methods,the multi-feature fusion method has better experimental results regardless of whether it is in the same database or across databases.Secondly,this thesis proposes an anti-spoofing detection algorithm based on ACE.The anti-spoofing detection method introduces the ACE algorithm to reduce the impact of light in the process of face anti-spoofing detection.The ACE algorithm is used to extract the features of real and fake faces and then classified by neural network.Experiments verify that the algorithm can also get better experimental results.Finally,this thesis also designs and implements a face-based student anti-spoofing detection system.The system mainly includes three modules: student account and password login,student face information storage,and student body detection and login.The account password login module mainly saves the basic account information of student users;the student face saving module is used to save the face information of student users;the student anti-spoofing detection login module matches the student's face and account information for system login.This module is mainly used an algorithm based on multi-feature fusion is used for face matching.The test of the system shows that the constructed system can effectively identify accessing student users,which lays a foundation for the future application of the log-in system for student life detection in actual scenes.