Research And Implement Of Real-time Face Recognition System In Video Surveillance

College entrance examination is an important means of selecting talents,for the purpose of ensuring fair and impartial examination,only authorized personnel are allowed to enter the workplace.In order to solve the admission work site admission personnel identification problem,this paper puts forward the use of video surveillance on the admission work site entrance,for the identification of the face in the video image capture in computer.To prevent the forgery of face image deception machine through automatic face recognition,it is necessary to determine whether the human face in the video is real face.Based on the relevant literature,this paper analyzes the change of the facial landmarks in the video and proposed normal orbital length ratio based face liveness detection method.By comparing the variance with the threshold value,we can monitor the blink behavior in the video,and then determine whether it is a real face.The method has the advantages of easy calculation,less memory occupation and no additional sensors,and the accuracy rate of more than 92.4% is obtained in the test.This paper researched on shape index feature based face detection and alignment,deep learning algorithm GoogLeNet based face recognition.Finally according to the above method using.NET,EmguCV and FaceSDK libraries on the VS2013 platform to complete a accurate and efficient video surveillance multi-face recognition system.The system meets the needs of the college entrance examination personnel identification,in 2016 in Henan province college entrance examination in the deployment of work,and achieved good practical application effect.