Real-time Face Recognition System Research And Implementation

This paper studies the popular face recognition technology, a variety of methods for face recognition, a real-time face recognition system. There are four stages in face recognition:the human face images for face detection and location of facial feature extraction and feature matching, to select speed, high precision algorithms through research and integration of computer vision field in the middle algorithm, the correction of the face pose and illumination compensation in the recognition process to improve the performance of real-time face recognition system.The major research work is as follows:1. Examining how to achieve real-time face recognition system. The face of real-time requirements, compare recognition of different algorithms for each step of the process, the recognition speed and high accuracy algorithms and achieve.2. Studied the localization algorithm based on of ASEF the human eye to understand the algorithm principle and training, the application mechanism. Face correction algorithm based on ASEF is designed to be adjusted, and complete correction of the face direction, different posture that may arise for the human face of the real-time face recognition system automatically adjusts to complete the face pose.3. Research the LBP feature facial feature extraction algorithm. And use the LBP feature based on the manifold LPP algorithm to reduce the dimensionality, LBP-the LPP characteristics of the image noise is even less sensitive, and can be extracted to a richer image of local and global information has a stronger representation of the face image and discriminant capacity, more robust, which can significantly improve the recognition rate.4. Completed the implementation of real-time face recognition system, entirely in C++algorithms, highly portable, modular design, so the whole system is an open platform, can easily replace and improve the algorithm.