The Optimization Of Multi-View Face Recognition Technology Based On SIFT Algorithm

With the progress of social technology and the improvement of people’s living standard, the problem of information security has been paid more and more attention. Therefore, a lot of identity authentication technologies appeared in recent years, of which the face recognition technology has rapidly became a research hotspot. The face recognition technology involves many areas such as image processing, computer vision, pattern recognition etc., and is widely used in the identity authentication, public security, criminal identification, human-computer interaction, intelligent access control system and other fields. The multi-view face recognition, which differs from frontal face recognition, is a difficult problem of face recognition technology. Because the tilt angle is unpredictable and facial rotation may increase the matching difficulty and reduce the matching rate. Therefore, the research area of the face recognition technology tends to the problem of multi-view.SIFT algorithm is a matching algorithm based on local feature, and has caused widespread concern for its high recognition rate and recognition speed. However, through the specific experiment we find that the matching rate of SIFT algorithm drops quickly in the condition of large facial illumination variation or large tilt angle. In order to improve the matching rate in the case of large facial illumination variation and large facial tilt angle, this paper has done the following research work: 1) Analyze the reason why the matching effect is poor in the condition of large facial illumination variation and large facial tilt angle from the generation method of SIFT feature descriptors.2) Propose a novel feature matching algorithm to apply the BP neural network to image matching. The basic idea is to take the successful-matched SIFT feature points as the training samples to establish a space mapping model based on BP neural network. Then, use this model to predict the possible coordinate of the SIFT feature point and search the possible matching points around the coordinate. The experiment results show that using the prediction model can reduce the number of mismatching points effectively and increase the number of correct matching pairs at the same time, which lead to the improvement of the matching rate of SIFT algorithm.3) Propose a novel recognition algorithm based on SIFT algorithm and feature matching, and get the higher recognition rate.