Research On Face Recognition Based On Single Sample

With the continuous development of society, face recognition technology has been widespread concern. Face recognition technology plays an important role in biological recognition, after 30 years of development, has made considerable progress. Currently most of the face recognition system can exhibit good recognition performance in the general environment, but most of this kind of face recognition systems require adequate face samples and appropriate application scenarios. With the widespread popularity of the use of second-generation ID cards and video surveillance, based on a single sample of face recognition technology has broad application prospects.AdaBoost face detection algorithm performs well when the picture is in small size, the performance is not satisfactory when the video image size is too large.In order to accelerate the speed of face detection, we use skin model AdaBoost algorithm, which can greatly reduce the detection range of the algorithm.Local Binary Pattern is a way to describe the texture. In recent years, researchers have been successfully applied to face recognition technology, and achieved remarkable results. Because LBP features operator has strong ability to describe the texture, while the calculation and implementation is relatively simple, making the LBP-based face recognition technology has made more research. Because Local Binary Pattern is a local texture operator, this paper presents an improved LBP operator, which can take care of the local features and dominant features. In order to further improve the ability to describe the face, this paper proposes a multi-level threshold LBP face recognition method can effectively single sample recognition.We use the proposed face recognition algorithm to conduct experiments, respectively, in the ORL face database, YALE face database, FERET face database and ID photos. Finally, we achieve a face recognition system based on the second generation ID card.