| With the development of machine learning,face recognition became more and more popular.We can identify or verify a person by face image.Face recognition is usually achieved by comparing the features of the test face and the features of face database.Although the face recognition technology has been studied for decades,is still an active topic.Because of developing a human face recognition system has a lot of requirements,such as accuracy,robustness,processing time and the limitation of image,so there are still many problems need to solve.The main content of this paper is the study of face recognition based on local feature.It is important to find an appropriate algorithm which can make the technology of face recognition timeless.The content of this paper is as follows.Firstly,the article briefly describes the research background,significance as well as domestic and foreign development facility application achievement of face recognition.Then,several typical methods of face recognition that based on global features and local features are introduced.The advantages and disadvantages of various methods in the light of the background,pose and expression are analyzed.What's more,this paper proposed a new feature point extraction method based on KAZE algorithm.The algorithm in dealing with illumination,scale and rotation is robust.But due to the high computational complexity,therefore it is unable to apply face recognition systems.For this,the improved algorithm of KAZE is proposed,that is a new feature extraction method based on AKAZE algorithm which reduced the characteristic point examination time to a great extent,and improved the face recognition rate of KAZE algorithm.At last,another method of face recognition based on local features is researched.That is a method which based on BOW and SURF local features.The algorithm not only has a high recognition rate and shorter calculation time,but also has strong robustness in the processing of fuzzy,compression and noise. |
PDF Full Text Request