Research On And Implementation Of Face Recognition Algorithm Based On SIFT

The development of science and technology,promote the rapid development of economy.It also puts forward new and higher requirements for identity recognition in public places.However,the traditional method of identity recognition through personal identification items and knowledge of personal identification has not been able to meet the needs of the society.After that,the biometric identification technology has been widely used in automatic recognition of human face images because of its advantages such as hard to be lost,unable to be copied and security.The face recognition system uses computer to detect,analyze and identify the image captured by the camera in the scene.According to that,it can determine the identity of the person in the image and realize the purpose of identifying the human face.This recognition method can be used to provide more secure,fast and efficient verification for the needs of authentication.So there are different kinds of face recognition algorithms proposed.Among them,the SIFT(Scale Invariant Feature Transform)proposed by Lowe David has been applied to face recognition research field because of its scale invariant rotation and robustness.Firstly,this paper did research on face detection based on improved YCbCr skin color model.Secondly,it also did research on face image feature extraction and registration,which is based on the algorithm of SIFT.The main research work is as follows:1.In view of the previous face detection and localization,a method of face detection based on YCbCr improved skin color model is proposed.Firstly,the skin color model is established to extract the skin color,and the face is selected through the area ratio and the length and width ratio of the face;Secondly,in order to remove the possible interference of the ear and neck regions,a ear localization algorithm and a method to locate the human chin based on the center line of pixel gray [CLPG] are proposed.The result of experiment shows that the method is accurate and effective.2.Since the number of feature points extracted by SIFT algorithm is too large and the matching is too difficult,this paper proposes an improved algorithm of SIFT to match the face image features.The irrelevant interference is removed by previous face detection method.It avoids the irrelevant feature point extraction and matching and ensures the validity and accuracy.3.At the stage of feature vector matching,the vectors are partitioned and matched.And then the minimum Euclidean distance is used as the standard of face features matching.At last,the ratio of the minimum Euclidean distance to the second minimum Euclidean distance is used to verify whether the feature points are valid matches.It can improve the matching efficiency and accuracy.