Face Recognition Algorithm Based On Multi-features Weighted Integration

Since the late sixties of last century,face recognition technology which uses computers to obtain effective facial features from face images and recognizes their identities has come into our vision.As an important application of artificial intelligence,pattern recognition and other fields,it attracts the attention of many researchers.Due to the continuous efforts and innovations of scientists and more convenient images acquisition,this technology is mature and has been applied to our lives,such as identity verification of bank cards,the company's attendance,security checks of airport or other areas and so on.Practical application shows that in-depth study of face recognition technology still has a very important value and significance.Current face recognition algorithms have quite good experimental results in ideal conditions.However,in practical applications,the stability of the face recognition system has been greatly challenged by the changing external environments(such as lighting)and the identified person's own influence factors(such as expressions,gestures,ornaments).Therefore,it requires researchers to develop a more stable,efficient and faster face recognition algorithm to eliminate the effects of these factors.Based on the above purpose,this paper presents a face recognition algorithm based on multi-features weighted fusion.Its main idea is first to extract different features from face images and construct the classifiers separately,and then to combine the multiple features classifiers using the weighted sum method in order to form a union face recognition classifier,finally using the union classifier for face recognition.The main works of this paper are as follows:1.A face recognition algorithm based on FisherFace and SIFT weighted integration is proposed.Firstly,FisherFace is extracted from the face image by Fisher linear discriminant analysis.Then,extracts SIFT feature points of the face image to construct SIFT local features using the K-Means algorithm.Local features are weighted using probability statistics method to compute their similarities.Finally,on the similarity level,Fisher Face global feature and SIFT local features are combined using the weighted sum method to construct a union classifier which is used for face recognition,weights of the features are determined by the method of probability statistics.2.A face recognition algorithm based on multi-features weighted integration is proposed.For the problem that the first algorithm has poor performance in face images with posture changed,firstly,use the affine transformation method to correct the postures of the face images.Then,extract Gabor feature as texture feature on the basis of the first algorithm.Finally,FisherFace global feature,SIFT local feature and Gabor texture feature are weighted using weighted sum method to construct a union classifier to improve the accuracy of face recognition.For the two algorithms above,compared experimental are designed on the CAS-PEAL face database and the experimental results are analyzed.The results show that the first algorithm has a better recognition performance for frontal face images with expression,illumination and occlusion influencing factors.The second algorithm has a certain increase in recognition rate for face image with posture changed.Therefore,our works in this paper have certain significance for the development of face recognition technology.