The Research On Liveness Face Detection Algorithm Based On Convolution Neural Network

With the rapid development of computer vision,the security authentication technology based on biometric feature recognition has become a hot topic.Among all kinds of biometric feature recognition techniques,face recognition technology has been widely used in various authentication systems because of its naturalness,non-contact and concealment,and the research on the security of face detection has become an urgent problem to be solved.Based on the theory of convolution neural network,this paper studies the feature extraction and classification decision in liveness face detection algorithm.The main content includes the following points:1)Aiming at the problems of traditional face detection methods,such as single feature extraction.and long training time,gradient easy to disappear and over-fitting based on deep learning algorithm,this paper proposes a novel face detection algorithm RFC-CNN based convolution neural network(CNN),CNN uses batch normalization(BN)method and multiple types of nolinear units to improve the algorithm detection performance.Explores the effect of different convolutional structures and different nonlinear elements on liveness face detection algorithms.A liveness detection algorithm DFC-CNN based on multi-type extraction is proposed.The experimental results show that the algorithm can classify the face images accurately2)In order to solve the illumination difference of different face images and increase the difference between the faces of authentic faces,the illumination compensation strategy is introduced into the preprocessing algorithm.Based on this,a design based on asymmetric convolution and greed is designed.The connexional convolutional neural network,through greedy feature extraction methods and shared network strategies,avoids feature omissions in the process of network learning.Experiments have shown that the algorithm can speed up the detection of live human faces while accelerating The convergence speed of the network model.3)In order to solve the technical complications of the existing face enhancement methods,the paper uses Gaussian curvature filter face enhancement preprocessing and designs an optimized convolutional neural network,in which CNN uses a composite parallel convolutional neural network and two mean values.The pooling strategy shows that the two-line parallel convolutional neural network face detection algorithm can accurately classify face images and increase the number of samples and training time.[The paper uses Gaussian curvature filter Face enhancement preprocessing means,and an optimized convolutional neural network is designed.CNN uses a composite parallel convolutional neural network and a two-mean pooling strategy.Experiments show that the two-line parallel convolutional neural network face detection algorithm can Accurately classify face images and increase the number of samples and training time...