Research On Color Face Recognition Based On Fusion Of Components Feature Extracted By Spectroface And FLD

As information and security technology develops, the traditional security and authentication technology has become increasingly difficult to guarantee the safety of certification. Biometrics recognition combines the information technology and biotechnology by using its unique physiological and behavioral characteristics for personal identification, and it is considered to be the highest level of security system in today's highly interconnected information society because of its characteristics that is difficult to copy and counterfeit.As a research hot spot in biometric recognition area, face recognition has broad application prospects and urgent practical demands. In addition to shape, texture and other low-level image features, the color information is a very important feature. The previous studies are mostly confined to gray images. We live in a colorful world, thus, the images that human visual system generates are mostly colorful. Previous research suggests that adding color characteristics with more information is crucial to further improve the identification system. Adding color information can effectively improve the recognition results, especially when the image characteristics (resolution, contrast, etc) recesses.In this paper, we study the human visual system and a variety of color models in detail, and then we put forward a new face recognition algorithm for color images—the algorithm based on fusion of components feature extracted by spectroface and FLD, which is based on the combination of gray-scale image recognition results and color image basic theory. Experiments show, compared with the traditional algorithms that based on quaternion model and color image graying, that the new algorithm has different degrees of advantages in complexity, the recognition speed and recognition rate, especially in the condition of large sample. Specifically, this paper mainly includes the following key points:1. Give an account of the research background, key technologies, development and present situation of biometrics and face recognition.2. Based on the deeply study of key technologies in face recognition area, such as basic theory of color image, color image preprocessing, feature extraction and classification, we propose a new algorithm based on fusion of components feature extracted by spectroface and FLD, and verify it by an experiment in a specific color space. In short:First of all, we preprocess the training and recognizing image by color image processing technology, and then extract the spectroface and FLD feature of the most effective components in color space, calculate the fusion feature by integration of the component feature, finally obtain the recognition result by the nearest neighbor classification algorithm.3. Summarize the characteristics of this new algorithm:First of all, from the view of realization method, we extract the face feature by combination of spectroface and FLD, which can effectively eliminate the effect of facial expression changes, light, side view. The feature extracted by combination of spectroface and FLD has a better classification property, fusion feature of components can better reflect the classification information of color image. Secondly, from the view of realization process, we give full consideration to the importance of color component for recognition, and select the most effective components—S and V to simplify the complexity, then merge the two component features according to global feature theory. The fusion feature enriches the classification information, enhances the anti-interference ability, and thereby improves the recognition rate.