Design And Implementation Of Video Image Face Recognition System Based On Deep Learning

With the development of society and the innovation of technology,people's demand for social security has been continuously strengthened.Among them,face recognition based on video surveillance is one of the key technologies.Video surveillance face recognition technology is more prominent in the current stage,such as image clarity,system real-time,complex scenes,etc.,with the emergence of deep learning technology has a certain degree of improvement.The thesis completes the system design of face recognition under video surveillance based on deep learning technology,and improves and optimizes the face recognition technology under video surveillance.The main contributions are as follows:(1)Based on MutiTask idea and MTCNN model,a cascade convolutional neural network model for face detection and classification suitable for video surveillance scenarios is designed.The network structure is optimized,the generation of face candidate regions is reduced,and the output of face matching scores is increased,and the final detection speed is updated by 55.7%.(2)Defined a face blur degree value suitable for face recognition system under video surveillance,which can improve the accuracy of face recognition according to it,and improve the image acquisition process for a certain frame of video stream to some extent.The resulting confusion,smear and other interference factors reduce the rate of misrecognition.(3)Based on the video surveillance scene application,based on the GOTURN visual tracking algorithm,the improvement and optimization are carried out,and the output of the search target confidence score is increased to better determine whether the target is successfully tracked.In the video surveillance scenario,combined with the face detection module,face detection and recognition can be performed more accurately.Based on the above research,the paper designs and implements a set of face detection and recognition system in video surveillance scene,realizes single face recognition system based on simulated “entry detection” scene;“security verification” scene Entrance Personnel Identification System;video recording face screening system for "Public Security Screening" scenarios.Tested in the actual environment,the indicators meet the functional requirements,verifying the effectiveness of the improved and optimized system.