Research On 3D Face Image Recognition Based On Deep Learning

In recent years,with the continuous development of computer technology,biometrics technology has attracted more and more attention,such as fingerprint recognition,iris recognition,face recognition and so on.Because of its direct,convenient and simple operation,face recognition technology has been widely used in online payment,security defense,criminal investigation and other fields.Face recognition is divided into two-dimensional face recognition and three-dimensional face recognition.The recognition technology of two-dimensional face image has been basically mature after years of development,but it is difficult to solve the problems of pose and illumination.The three-dimensional face data has rich spatial information of face,which has become a research hotspot.At the same time,the deep neural network is designed to extract features automatically in the case that the traditional recognition methods may affect the experimental results.The main work of this paper includes the following parts:(1)Face data preprocessing,firstly,the three-dimensional face point cloud data is transformed into depth map,and the depth map is processed by face detection,segmentation and normalization.At the same time,LBP features are extracted from depth map to improve the reliability of data in experiments.(2)A deep learning network is designed to extract convolutional features of deep face images and their LBP feature maps respectively.In the feature extraction part,the depth maps containing face depth information and LBP features that can effectively describe the characteristics of depth maps are used.The convolutional neural network is used to extract features automatically,and the neural network is used to extract features while selecting effective features.The feature output layer acts as the input of the recognition network,realizes feature fusion,and combines the improved convolutional neural network with soft-max classifier to complete classification and recognition.(3)The three-dimensional face recognition system designed in this paper is tested,and the depth images and LBP features of face are experimented respectively.At the same time,the single feature experiments are compared.The results show that the feature fusion method adopted in this paper has good stability.Compared with the single feature recognition method,the method designed in this paper has a significant improvement in the accuracy of recognition.(1)Face data preprocessing,in view of the difficulty of processing the three-dimensional point cloud data,first converts the threedimensional face point cloud data into the form of depth map,and then carries out face detection,segmentation and normalization on the depth map,in order to eliminate the irrelevant information in the image and provide a good data source for subsequent experiments.