Research of Face Recognition Based on2D-Gabor Feature and Multi-feature Fusion
|Keywords||Face Recognition 2D-Gabor Feature SVD Feature fusion|
Human face is the most common mode of human visual and is also one of the most important human biological characteristics. Knowing a man, the main is to remember one’s face. Through the analysis of human face, we can get a lot of important information related, such as age, gender, expression, race, etc. In the21st century, with the rapid development of computer hardware and software, computer vision and pattern recognition in image processing become widely known as a new force. Recently years, face recognition also has quietly become a hot topic in the research area in computational intelligence and pattern recognition direction’s researchers all around the world. Whether in theory or in the actual, face recognition has made a good development. Especially in information security, smart card system, human-computer interaction, monitoring, intelligent traffic has more broad application prospects. At present, there has been much system using face recognition technology in user identification such as the escaped criminal’s identification of public security system, automatic check of the motor vehicle driver’s license and entry and exit passport, government agencies who has higher security demand, the automatic identification system of the bank, various automatic monitoring system, laptop and cell phones and other electronic product’s user identity, etc. Compared with other biological characteristics such as human fingerprint identification, iris identification, speech recognition, face recognition is more direct, convenient and friendly.In recent years, the artificial intelligence and pattern recognition researchers around the world have done a lot of research and put forward a lot of excellent algorithm. Face recognition technology is developing very rapidly, and has obtained great achievements, but there are still a lot of problems such as complex background image, illumination, the point of view of face image, and so on. Human face is a kind of rigid body and features are not easily accurately described. Because of the growth as the age will produce some slight changes such as wrinkles, freckles and so on. In addition, the face can also have many other adornment objects such as such as beard, eyelashes, glasses, peak and so on. The light intensity and angle of the face shooting are difficult to accurately control, changing constantly, which have been difficult problems in face recognition technology to be solved urgently. How to quickly, accurately identify face is a hot topic in the research field during more than ten years.In this paper, propose a method based on the fusion of2D-Gabor feature and singular value decompose feature. Because2D-Gabor feature has good robustness on the light intensity and change of face angle, so2D-Gabor feature in the feature extraction and face recognition field obtained very extensive application. In addition,2D-Gabor feature also have very good local feature and direction of space selective feature and it can also accurately extract the complete texture feature of a picture. So, many of the existing face recognition algorithms use the2D-Gabor feature as a human face recognition feature.2D-Gabor has taken possess of a place even has become an indispensable method of feature extraction algorithms in face recognition.On the other hand, image singular value features (SVD) has a very good stability. Singular value reflect the characteristics of image data a kind of intuitive algebra essence. Singular value decomposition reflect a non intuitive algebra essence characteristics of image data. Singular value decomposition feature have constant algebra and geometry characteristic. Singular value decomposition feature is a kind of effective algebra method of feature extraction and it is widely used in image data compression, signal processing, pattern recognition, etc. In this paper, we get the face of2D-Gabor characteristics and singular value decomposition characteristics first and then fuse them together, to make sure higher differentiate. Using of the fusion feature to train samples and then using the nearest neighbor classifier for face classification. The test results in Yale and on ORL face database show that compared with the method of only based on SVD or use only Gabor feature, this paper improve a lot.