Research And Application Of Facial Iris Dual Feature Fusion Recognition Algorithm Based On Deep Learning

The biometric identification method uses the unique physiological or behavioral characteristics of individuals such as fingerprints,iris,face and voiceprint to realize identity authentication,which is more secure,convenient and reliable than traditional methods such as passwords and IC cards.Single biometric identification is easily affected by environmental noise and spoofing attacks,which can not meet the requirements of high security occasions.The fusion of multiple biometrics has a more stable identity expression ability,providing a more effective solution for identification.A stable,efficient and accurate fusion method is the key to multimodal biometric recognition.The human face and iris are in a compact position,making it easier for acquisition equipment to acquire them at the same time.With the rapid development of deep learning technology in the field of biometric recognition,multimodal biometric fusion recognition opens up a new research direction.Therefore,under the framework of deep learning,this paper studies the face and iris dual-modal fusion recognition algorithm and model.The main research work of this paper is as follows:1.Aiming at the lack of real multimodal data sets of face and iris,we collected and established a multimodal data set of face and iris,which laid a data foundation for subsequent research work.2.Based on the YOLOv3 model,a fast iris localization detection algorithm is proposed.The iris localization problem is transformed into a target detection problem,and the traditional iris localization method solves the problems of low localization accuracy,high computational complexity and poor robustness of iris images with poor quality in complex environments such as specular reflection and blinking.3.Convolutional neural network can automatically learn the advantages of high-level and abstract feature expression of images,conduct research on face and iris single-modal recognition algorithms,and design Shuffle-FaceNet face features based on improved attention mechanism.Extraction network and MFARM-IrisNet iris feature extraction network.4.The algorithm and model of facial and iris dual-modal feature fusion recognition are studied.In the feature layer,two fusion methods of face and iris features are studied,and the two modal feature vectors are analyzed by the adaptive weighted fusion method and the low-rank multi-modal feature fusion method based on the modality-specific low-rank factor.The construction of feature fusion recognition model.The experimental results show that the dual-modal feature fusion recognition has better performance than the single-modal recognition.5.A dual-modal fusion recognition model based on transfer learning is designed,which solves the problem that it is difficult to train accurate models with small sample data sets.6.Based on the face and iris fusion algorithm,the face and iris fusion identification system is designed and implemented.