Font Size: a A A

Research On Multi-pose Face Recognition Based On,Single View

Posted on:2017-01-11Degree:MasterType:Thesis
Country:ChinaCandidate:R Y MoFull Text:PDF
GTID:2348330536976743Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Face recognition has significant commercial value in the field of public security,digital identity and multimedia because of its characteristics of high precision and non-contact.At present,many face recognition algorithms can achieve good performance under the condition of controllable and cooperative.However,there are many challenges for face recognition technologies while dealing with multi-pose and single view problems.This thesis study on single view and multi-pose face recognition problems systematically.There are the following two main researches in this thesis.1.A novel method based on the combination of linear regression algorithm and support vector machine is proposed to solve the problem of pose-invariant face recognition from only one image.Firstly,facial feature points are located based on view-based AAM(Active Appearance Model)and face images are aligned;secondly,mapping from the non-frontal image to the frontal image is constructed based on the algorithm of linear regression and frontal faces are obtained from non-frontal faces with different poses;thirdly,SVM(Support Vector Machine)is used to classify the facial features.Experimental results based on the CAS-PEAL-R1 face database show that the recognition rate is 86%,which is better than other approaches for pose-invariant face recognition.2.A 3D face reconstruction method is based on a single image proposed to solve the lacking of training samples.Firstly,valid face region is segmented automatically and planar template is defined by the segmented faces;secondly,a 3D sparse morphable model is established and 3D face can be reconstructed.thirdly,virtual multi-pose face images can be obtained by texture mapping,rotation and projection of the 3D face model and training samples are enriched.Then,BP neural network is used for classification.Experimental results based on CAS-PEAL-R1 face database show that performance of the proposed approach is 91%,which is better than other approaches for pose-invariant face recognition based on single image.3.On the basis of the above research,a face recognition system based on Matlab is designed and implemented.This system contain the two methods mentioned in this thesis.
Keywords/Search Tags:Face Recognition, Linear Regression Algorithm, Genetic Algorithm, Support Vector Machine, 3D Face Reconstruction, 3D face sparse morphable model
PDF Full Text Request
Related items