Font Size: a A A

Video-based Real-time Face Recognition Research And Practice

Posted on:2011-02-01Degree:MasterType:Thesis
Country:ChinaCandidate:M XuFull Text:PDF
GTID:2208360305493551Subject:Biomedical engineering
Abstract/Summary:PDF Full Text Request
Biometrics identification technology has leapt forward in recent years. Its prospect of application is very extensive. Compared with the traditional method of identification, biometrics identification technology is safer, more secretive and convenient. It is hard to be forgot and stolen, it also has the advantages of good anti-counterfeiting property and convenience of taking along and be used anytime and anywhere. especially fingerprinting, palmprint recognition has impregnated into many industry. But there methods have common shortness, it must have contrived contact, otherwise it's hard to identify. Contrast to other methods, face identification is direct, friendly and convenient, and it's easy to be accepted by the user. So, this technology has been studied and applied abroad recently.This paper aims at the status of face identification technology, analyzing kinds of face detect, face feature extraction and face identification arithmetic and comparing their capabilities (especially the real time), and then putting forward a real-time video based face recognition method:Adaboost+Haar+Cascad was used to detect the face in complex background, improved principle component analysis was used to extract the face feature and BP neural network was used to real-time face classification and recognition. On the basis of this, we achieve a real-time face identification system. The experiment proved our system has good stability, real time and high face recognition rate. The major part of this paper is as follows:(1) We use DirectShow technology to program video capture modules and capture video to image through camera, the frame rate is about 60 and it achieves the purpose of real-time display.(2) Eliminating light and normanizing for the detected face image, so that it's convinent for feature extraction.(3) A face sorter is trained and we implement a face detect system based on Adaboost sorter. This system is able to detect face in complex background. (4) Face feature extraction is achieved by the improved principle component analysis and face reconstruct validates the advanced of the method. The improved method also makes the expression of face feature more convenient and efficient and it supplies possibility for face recognition using BP neural network. (5) BP neural network recognize face. A real-time face recognition platform is established preliminarily.Adaboost (2D)2PCA and BP neural network real-time face recognition method is not only advanced in arithmetic, but also easy to realize. The system is realized in VC++6.0 platforms and it has the advantages of stability, real-time and practicality.
Keywords/Search Tags:real-time video, biometrics feature, face recognition, Adaboost, principle component analysis, BP neural network
PDF Full Text Request
Related items