Design Of Face Recognition System Based On Web Camera

With the development of national computer technology and chip technology,many embedded terminal products have appeared around us and played certain roles to complete specific tasks for people.For the application requirement of home security systems,this paper designs a set of face recognition system for webcam based on Hisilicon platform.The system consists of three parts: hardware module,software platform and functional implementation module.In the hardware module,the system uses Huawei Hi Silicon Hi3518 E chip as the central processor and the image acquisition uses SC2235 image sensor.The specific circuits include: Hi3518 minimum system circuit,audio acquisition and audio playback circuit,video acquisition circuit,storage circuit,wireless module circuit,IRCUT circuit design.The software platform includes the startup program,kernel,and file system.Uboot is used as the startup program,which supports Nor Flash boot mode and completes the system initialization work at system power-on;Tailored kennel configured system hardware drivers,and manages and allocates system resources;Using the JFFS2 file system,the file system is transplanted using busybox,which matches the kernel to achieve the orderly management of file.The functional implementation module is achieved by five modules,which are image acquisition,preprocessing,face detection,face live detection and face recognition.The image acquisition uses the MPP provided by Hi Silicon.The system is initialized first,then the video input module parameters and the VI channel binding relationship are set.The system starts to capture images and store them in DDR,using memory mapping to implement image frame reading;The histogram equalization and filter are used to improve the reliability of the extracted features in the preprocessing module.The face detection module adds two Haar templates to the original Haar template to extract image features and use Adaboost to train a classifier,which reducing the false detection and miss detection of the side face;The face live detection uses Co ALBP and LPQ features to extract texture features in HSV and YCb Cr.The extracted fusion texture features are trained the SVM classifier,which can effectively recognize real and fake face;The face recognition module extracts LBP features using space conversion method to reduce the data dimensionality,which accelerates the speed of face recognition on the premise of ensuring the recognition rate.Finally,the system was experimentally verified.The verification results show that the system can use the camera to capture images for face detection,live detection and face recognition.For live detection detection tests,the recognition rate is over 96%,and the correct rate for the face rercognition can reach 93%.The system takes about 2.3s from image acquisition to face detection,live detection and face recognition.The realization of this system proves the feasibility of the face recognition system based on Hi Sicon webcam,and has certain practical value.