RETRACTED ARTICLE: Video Face Detection Based on Deep Learning

In recent years, with the development of Internet plus concept, online identity has become a major problem based on the continuous expansion of network applications. The online authentication technology based on biometric features can maintain the consistency of human digital identity and physical identity, so people pay more attention to it. This paper studies the problem of human face detection. The main tasks are as follows: an active body detection algorithm for convolution neural networks based on dynamic feature is proposed. First, the Pyramid LK optical flow method is used to track the video, and the dynamic information of the image is obtained. Then, the information of the optical flow is analysed, and the horizontal and vertical displacement are calculated. According to the two displacements, the displacement amplitude diagram is calculated, that is, the dynamic feature graph. The dynamic feature graph is used as the input of the convolution neural network. Finally, the feature extraction and the living detection are carried out. A face authentication system with living face detection function is designed. The system includes the registration phase and the authentication phase. The registration phase includes face image detection and feature extraction module. The authentication phase includes face detection, human face discrimination, feature extraction and similarity calculation module.