Research On Key Technologies Of Face Recognition Based On Android

Recently,biometric recognition technology has been widely used in daily life,such as Iris recognition,facial recognition,fingerprint identification and so on.With the rapid development of face recognition technology,it also has good stability in dealing with posture,illumination and expression changes.Because of facial recognition,with high efficiency and convenience,has been widely used in various scenarios,such as access control and log in systems.However,the face recognition system can only recognize the face image at present,but can't judge whether the collected face image is from a real person or a captured photo.Therefore,it is extremely vulnerable.With more kinds of spoofing,like using printing photos and Gif,face anti-counterfeiting detection attracts more and more attention from researchers.In this paper,the face detection method based on MTCNN is used to realize real-time detection,compared with the traditional face detection method,the detection effect is better.At the same time,based on the further realization of convolutional neural network to achieve face recognition,the face anti-spoofing detection module is designed and implemented.Different from the method of single facial feature used in existed face image,a method combining two features and utilizing Convolutional Neural Network to learn face security feature was proposed in this paper.The scheme is tested in two public databases,in which the accuracy of the anti-spoofing on the CASIA database reached 94%,and the accuracy of the anti-spoofing on the REPLAY-ATTACK database reached 88%.The result shows that this method can be used to discriminate whether the face image belongs to real or fabricated face much more efficiently,compared with the texture feature based on DOG image extraction and based on image distortion analysis,etc.the scheme has better performance in Anti-face spoofing.Finally,this scheme is implemented on the Android platform and works well.