Research On Face Recognition Method Based On Global Feature And Local Feature

With the rapid development of pattern recognition, computer vision and image processing technology in the information society, the face recognition as an important means of authentication has been concerned generally by the society. Face recognition is acclaimed because of its uniqueness, safety, convenience, effectiveness, non-contact, and not easily detected and many other features. Recently, commercial products of face recognition are more and more, such as most airports, banks, ports and other important places, where have been deployed and installed face recognition systems. Face recognition plays an important role in our daily life. In the face recognition processing, image acquisition may be affected by the changes of light, viewing angles, facial expression, age and blocking, so the image of the same face may appear large differences, which result in the difficult of identification. Therefore, improving the recognition accuracy of face recognition systems is one of the important goals in the study of the face recognition technology.Firstly this paper introduces the subject background and significance of the face recognition, introduces the basics, the relevant theory and technology which is relevant to the topic, then describes the local feature based on the face recognition technology and overall characteristics, and introduces the face recognition technology which is most widely used based on scale invariant feature (Scale Invariant Feature Transform, SIFT) and kernel principal component analysis (Kernel Principal Component Analysis, KPCA), and proposed extraction method of local features and global features of the experiment in the paper. When extracting local features, the paper establishes the coordinate system of the human faces and uses14key feature points as the local feature points according to the aesthetic analysis report which can be used in the face recognition. When extracting the global features, the paper uses the distance between the features, angle characteristics, perimeter characteristics, area characteristics, characteristics of size and facial proportions characteristics and so on as the global features, and when obtaining feature vector groups, which can be used in face recognition. In the end,the paper used six classic face images in the AR face database to do the test for different expression characteristics of the local features and the global features.Finally, the improved D-S evidence theory is adopted to integrate the local features and the global features and then complete face recognition. The experimental system is completed with Microsoft Visual Studio C++6.0, OpenCV basic function library is adopted. Because the SIFT algorithm and the KPCA algorithm is applied more widely, the system compares the algorithm of the paper with the two algorithms above on AR face database and JDL face database. The experimental results show that the algorithm in the paper has a higher recognition rate.