Research On Local Occlusion Face Recognition Technology Under Deep Learning

In recent years,with the rapid development of deep learning,significant progress has been made in the field of computer vision.Face recognition technology,as an important branch of computer vision,has become one of the hot spots of current research and has a wide range of application prospects.Although face recognition technology in specific environments has matured,the accuracy cannot be guaranteed in the case of local occlusion.As facial occlusion may lead to the loss of key facial features,coupled with the complexity and similarity of facial feature structures,it leads to poor recognition accuracy performance.Therefore,the research on face recognition with partial occlusion is very urgent,and the main contents and results of this paper for the problem of face recognition with partial occlusion are as follows:(1)To address the problem of local occlusion face image restoration,this paper proposes a local occlusion face image restoration algorithm based on Generative Adversarial Networks(GAN)to realize the restoration task of local occlusion face images.This paper adds a U-Net network to the original encoder-decoder,inputs the detailed feature information from the encoder to the corresponding layer of the decoder through jump connections,and introduces a total variation loss function to reduce noise and other unwanted high-frequency details,which is closer to the original image and improves the restoration effect of the occluded image,thus making the restored image smoother and more natural.The quality of the restored image is evaluated by two evaluation indexes,Peak Signal-to-Noise Ratio(PSNR)and Structural Similarity(SSIM),which objectively show that the improved partial occlusion face image restoration algorithm in this paper is effective.(2)In order to enhance the utilization of feature information such as shallow texture and details of face images,a pyramidal multi-scale feature fusion network is added to the Res Net50 face recognition network to effectively fuse the deep features extracted by the Res Net50 network with the shallow features,thus improving the accuracy of partially occluded face recognition.The interference of the occluded region is mitigated by the image restoration algorithm,and then the images generated by the image restoration algorithm are recognized using the improved local occlusion face recognition algorithm.The experimental results show that the improved local occlusion face recognition algorithm is used to recognize the repaired occlusion images under the occlusion area size of 10%,20%,30%,and 40%,respectively,and the accuracy of the improved algorithm in this paper is improved by 4.62%,4.95%,3.9%,and 3.9%,respectively,when compared with the Res Net50 face recognition network by fusing deep and shallow features,8.03%.