Research On Occlusion Face Recognition Algorithm Based On Deep Neural Network

The vigorous rise of deep learning has caused a research boom in the field of artificial intelligence.As an important branch of the field of artificial intelligence,face recognition technology has made continuous progress in recent years and is widely used in business activities and people's lives.With the advancement in the field of artificial intelligence,the accuracy and accuracy of face recognition have been significantly improved.However,in real scenes,due to the influence of uncontrolled factors,the face is occluded,which leads to the lack of features of the face target,and ultimately leads to a sharp decline in the accuracy of the face recognition algorithm system.Based on the face occlusion problem,researchers have proposed various algorithms to solve the occlusion problem in real scenes,but these algorithms have not fundamentally solved the problem of missing features.Therefore,under natural conditions,identity recognition based on blocking faces is still a hot topic in the field of face recognition.This paper focuses on the occlusion problem of human faces,and proposes an improved generative adversarial network(GAN)to repair the missing faces in the article.It can fundamentally solve the natural problems by repairing and reconstructing the missing features of the faces.Under the conditions,the interference of the obstruction to the recognition,the experimental comparison proves that the proposed algorithm has a higher accuracy for the recognition of the face that exists under natural conditions.The main research work of this article is as follows:(1)Research on face detection technology based on traditional detection algorithms and deep learning algorithms.Based on the traditional face detection algorithm,the Adaboost integration algorithm is mainly introduced.Face detection algorithm based on deep learning,this article mainly conducts in-depth research on MTCNN face detection algorithm.MTCNN face detection algorithm is based on the commonly used algorithm in deep learning face detection algorithm,which can achieve face detection and key point positioning at the same time.Then,according to the position of key points and the distribution characteristics of facial features,the key area of the target face is located.The research on related technologies of face detection provides a sufficient theoretical basis for the subsequent face recognition and the algorithm to achieve face occlusion.(2)Research on face recognition based on CNN network.This article first introduces the basic principles and structure of the convolutional network(CNN),on which basis the CNN network is explained in detail,and then this paper proposes an improved VGGNet network to achieve face recognition.The method verifies the pros and cons of the face recognition algorithm of this algorithm and the VGG network on the relevant face recognition data set.The experimental results show that the improved VGGNet network has a better effect on the accuracy of face recognition.(3)Based on the effect of occlusion under natural conditions on the accuracy of face recognition network,this paper adopts the method of repairing the face occlusion area in detail to achieve the idea of improving recognition accuracy in face recognition algorithm.Based on this idea,this paper mainly proposes a local occlusion face repair algorithm based on W-GAN network to repair the occluded part of the face by studying related algorithms of face generation based on Generative Adversarial Network(GAN).Based on the purpose of repairing the details of face images,this paper introduces a new loss function on the face repair algorithm,and finally combines an improved face repair algorithm and a face recognition algorithm based on the improved VGGNet network to design a natural condition Under the face recognition algorithm,this paper conducts two sets of simulation experiments on the face recognition related data sets.The experiments show that the algorithm has good recognition accuracy for face occlusion.This paper has carried out related experiments on the above-mentioned algorithm.The experimental results show that the algorithm proposed in this paper has higher robustness to face recognition than the mainstream CNN-based face recognition algorithm in recent years.Especially,it has obvious advantages in recognition effect under occlusion conditions.