Font Size: a A A

Research On Face Recognition Method Based On Principal Component Analysis And Multi-core Support Vector Machine

Posted on:2019-03-31Degree:MasterType:Thesis
Country:ChinaCandidate:M W ChiFull Text:PDF
GTID:2428330545986631Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
In today's society,people pay more and more attention to the development of safe and intelligent urban space.Because face has natural biological characteristics,and has the advantages of easy access and not easy to copy,it makes face recognition in various fields.In practical application,face recognition requires large face database,how to improve the recognition rate of face in small sample cases,and how to face image feature extraction in the complex factors,better application in city monitoring system is the main purpose of the thesis.In this paper,in order to increase the face contour and improve the recognition of traditional single kernel support vector machine rate for the purpose,from the face image detection,face image feature extraction and face recognition analysis of three aspects of research on face recognition system,the main research contents of this paper summarized below:First of all,face image detection: aiming at the time consuming problem of face shape detection based on the active shape model,the face is detected by Haar-like feature.Then,the active contour model is applied to detect the contour of the human face,which improves the detection efficiency of the contour feature of the face image.Secondly,the feature extraction of face image: according to local two value model algorithm to extract the image features of high dimension sample sample,increase support in the calculation of face image classification and support vector machines,this paper selects the principal component analysis method for face image feature extraction,the method can make up for the local two value model algorithm problems.In view of the traditional principal component analysis algorithm will miss some contour feature information,on the basis of improved fusion of face contour feature,proposes a multi feature based on principal component analysis method,this method can effectively improve the face recognition rate under complex conditions,robustness.Finally,face recognition: contribution for different kernel functions are different,this paper selects the Gauss kernel function(RBF)of the local and global polynomial kernel(POLY)to form a multi kernel support vector machine,the kernel function is linear additivity with different weights,constructing multiple kernel learning the model,this method can increase the recognition rate of face image,improve the generalization performance of the sample.
Keywords/Search Tags:Face recognition, principal component analysis, contour feature, multicore support vector machine
PDF Full Text Request
Related items