Fusion Of Global And Local Feature For Face Recognition

Face recognition has always been a hot issue in the study of biometrics recognition technology. So far. face recognition research has yielded some results in the ideal state, however, face recognition under interference factors existing state is still no better solution proposed. Interfering factors include:illumination, pose variation, occlusion material, expression changes, etc. Researchers have proposed facial feature extraction methods based the analysis of local feature are useful to improvement of the existing face images of light and gesture. From this perspective, this paper studies the face local feature extraction and recognition, hen we effectively integrate the global and local features, thereby improving the robustness of face recognition algorithm, main tasks are as follows:(1) The history and current status of the face recognition, especially for current research status. This paper has done a detailed overview from the current university which is studying face recognition to face recognition products manufacturers, then we made a summary of the current face recognition market distribution and development situation of the future.(2) Adaboost method was used for face detection. Face region could extract from natural images, then face recognition and identification. In this paper, the main current face detection methods were outlined, then the adaboost algorithm and theory were discussed. Finally we show the effect of face detection.(3) The global feature extraction method based on subspace was studied. First we introduced the principal component analysis of facial feature extraction, then focused on two-dimensional principal component analysis based global feature extraction method.(4) The local feature extraction and face recognition based on Gabor transform were studied. We used a combination of block technology and Gabor feature extraction method for obtaining face structural information, and the two methods were proposed in block thinking, Identification process used the fuzzy classification decision rules. Finally, because Gabor transform and identify have a large amount of calculation, so this sampling technique was applied to the Gabor block face recognition, we achieved better results.(5) The integration of the global features and local features. We discussed the need of the feature fusion. Finally, the global features and local features based on the Gabor transform were effectively integrated, which improved the performance of face recognition system.Through the above work, we constructed the recognition system based on fusion of global and local features, experiments show that the ideas to improve face recognition system both from the perspective of global and local are effective.