Research On Face Recognition Algorithms With Age Change

In recent years,with the popularity of intelligent human-computer interaction devices,image-based face verification technology has attracted wide attention from all walks of life,and it has also made new progress and breakthroughs.The accuracy of face verification is often affected by illumination,posture,expression and age change,among which age change is particularly complex,such as the change of facial shape,skin color and wrinkles,which are the problems to be solved by face verification technology under age change.In addition,the research on face verification under age change is comparable.There is less attention to other factors,but with the increasing demand for face verification technology in various fields,it is very important to study the face verification method under age change.In this paper,a new face verification network model is designed for the above problems.Its main work and contributions are as follows:(1)This paper chooses CACD2000 cross-age face database as training sample of face verification network.However,the database contains not only face information,but also a large number of information unrelated to facial features.Firstly,the training samples are screened,and then face detection,alignment,tailoring and size normalization are carried out with the help of MTCNN network model.(2)By analyzing the performance of deep convolution neural network in feature extraction and image matching,this paper designs a R-S face verification network model which integrates residual network and Siamese network.The residual network structure is used as the basic unit of feature extractor,and the image matching function is realized by combining the Siamese network framework.(3)Cosine similarity is used instead of Euclidean distance in the original loss function to represent the difference of eigen vectors.By comparing and analyzing the experimental results,the accuracy of the network model based on cosine similarity loss function is improved by 1%.(4)The R-S face verification network model is tested and validated in this paper.Compared with most existing verification methods,the comprehensive performance of the model is significantly improved,and the accuracy rate is more than 95%.