Research On Age Prediction Method Based On The Combination Of Local Facial Features And Global Facial Features

In recent years,as an emerging biometric recognition technology,age prediction has a large application market in daily life,such as user portrait,market analysis,etc.Therefore,age prediction from facial images has become an important research direction in the field of computer vision.The face contains many unique features that distinguish individuals,but aging is a gradual process.The aging changes of the face will vary with different growth periods,different ages,and different external environments such as lighting background,and these facial changes tend to occur only in certain areas of the face.Therefore,There is no way to deal with these local region variations well when extracting global facial features from facial images.In addition,facial images contain more or less noise,which also increases the difficulty of age prediction in different degrees.In previous studies,there are few literatures that fuse local features and global features for facial age prediction.Therefore,this paper attempts to fuse the local facial features on the basis of the global facial features for age prediction research,expecting to further improve the prediction accuracy of age categories.In this paper,a quantitative analysis of facial features is carried out to study the trend of facial features during the aging process of individuals.In this paper,the experimental research is carried out on the FGNET facial data set.Firstly,the image is preprocessed to crop the whole facial image and the local facial features image,and then the different facial components are divided into blocks,and the local binary pattern feature extraction is carried out on the sub-regions of each facial component after division.The facial feature differences of the same age group and the feature differences between different ages after the feature fusion of each facial region were analyzed.This paper also proposes an age prediction method based on the fusion of five senses and facial global features,which can enhance the effectiveness of age prediction to some extent.This method considers the influence of the global information of the face and the change of facial features in the local facial regions during the individual aging process on age prediction,and designs a depthwise separable gated convolutional neural network model to fuse the features of different regions of the facial image to obtain richer texture information.In this method,the texture features and geometric features of each region of the face processed by the gated convolutional neural network model and the global key facial features extracted by the convolutional autoencoder model are fused,and the fused features are used for feature selection.Then the effective information of these features obtained by the fully connected neural network model and the support vector machine model is used for age category prediction.In this paper,experimental evaluation on FGNET,UTKFace,CACD,AFAD and Fairface datasets is carried out to verify the effectiveness of the proposed age prediction method.