The Study Of Face Recognition System Based On Monitoring Equipment

With the development of the society,the security of authentication is becoming more and more important.Biometric has advanced a lot in recent years.Among those technologies,the face recognition technology(FRT)has become one of the most important fields of the computer vision and the pattern recognition.Early face recognition research was mainly based on face images with strong constraints such as simple background,the same size of face,no tilt and so on.However,the actual monitoring environment is always very complex,the background is messy,the size of faces are different.Generally speaking,the main work in this paper is just as follow:(1)I studied classical face detection algorithms including skin model,Adaboost and Faster R-CNN.Through the experiments I decided to use Faster R-CNN as face detection algorithm in this system.(2)The original image passed by camera is RGB image.The high-dimensional data will not only wast storage space,and also affect the efficiency of the algorithm.So I converted the original images to a grayscale images.(3)I made reseach on the classic algorithm for face recognition: SRC.Experiment shows that the SRC algorithm has strong robustness to the change of illumination and partial occlusion.I compared PCA(Principal Component Analysis),LBP(Local Binary Pattern)and LDA(Linear Discriminant Analysis)through experiments.Finally I chose SRC + LBP as the face recognition algorithm of this system.(4)I rewrote some functions of the camera SDK(Software Development Kit)to get the frame and change it into the format that OpenCV support.(5)I designed a real-time face recognition system and achieve it by the way of OpenCV.First,the Faster R-CNN is used to detect and locate face in the video frame.Then,I use the SRC as the face recognition algorithm.Experiments show that the system deal with real-time video and has robustness to the illumination.