Research On Anti-Spoofing Methods In Face Recognition

Face recognition technology is widely used in identity authentication system. The spoof on face challenge the system due to the photos or videos acquired easily and the quality became higher. Therefore, the discipline of checking the face alive has became an interesting field, and it is also vital to check the liveness before face recognition.The method for mainstream of face liveness detection was introduced detail and the performance of those was compared in the article. The reason for higher accuracy and the time in every part were analyzed in face recognition based on background context and blink method, and the performance and the practicability were improved obviously reforming on this method. At last, the system of face liveness detection was established. The hardware and software were designed to achieve the function of distinguishing the living face.A method based on "Pyramidal Implementation of the Lucas Kanade Feature Tracker Description of the algorithm" to track the corners transforming was put forward. The time is only 6% of the former producing, which has shorter time and error.A way based on facial coordinate positioning and detecting eyeball by morphological operation was given as the method to judge alive by blink has a low accuracy. In the part of locating eye region, it improved the correct rate of eye region and had a shorter time to find the area by using facial coordinate positioning. In the part of judging blink, morphological operation to find the eyeball, analyzing the dynamic change of the pixels, can improve the speed to detection alive than the former using classifier, those can adapt to the requirement of real-time system.This face liveness detection system was constructed by VC++ and OpenCV. It is made up with image acquisition and image algorithm and the software interface. The result shows that it satisfied the requirement of the real-time and the robust for the environment, and had a higher accuracy.