Near Infrared And Visible Face Image Fusion Recognition Algorithm Based On Tensor CCA

With the development of computer technology and biomedical engineering andtechnology are advancing by leaps and bounds, the use of biometric identification identitybiometric technology more and more people of all ages. Biometric identificationtechnology is a combination of new information technology and biotechnology recognitiontechnology, by using a variety of equipment for the body’s physiological characteristics orbehavioral traits for identity authentication, biometric identification is an important patternrecognition component. All based on biometric identification technology has substantiallythe same work with the principles and processes. The first to use optical scanning sensor toobtain images of the human body, and its data, and then use a mathematical algorithm forprocessing the extracted image will be the last generation feature template with pre-storedin the biometric database to compare comparison Similar to determine whether they match.Face recognition technology is a biometric identification in the field of physiologicalcharacteristics based on the recognition technology, is based on face feature extraction bycomputer, a technology and authentication based on these characteristics. Compared withother biometrics, face recognition technology has the advantages of simple operation,intuitive result, good concealment.Due to the advances in optical technology, we can obtain the near infrared face imageis not visible to the human eye wavelength. The near infrared face image is not affected byambient light images, can even at night or no visible light conditions can be achieved.Because of the different ways of obtaining and visible face images, said the face differentinformation, so we can feature fusion of the two. Merging their feature recognition is calledheterogeneous face biometrics. So what can we use to carry on the fusion?Canonical correlation analysis of face recognition is an important research subject ofmultivariate statistical analysis. Firstly, its mathematics model, and the solution algorithm.With the idea of principal component, with a few of the comprehensive variable to reflectthe correlation properties between two group variables. The two set of variables in thecanonical correlation analysis is the need for the two sets of variables from the same sampleby different ways, that is the need for the two sets of variables has a certain relationship, just can be used in near infrared and visible face image recognition.The canonical correlation analysis can be used for near-infrared and visible light faceimage correlation calculation, the calculated two sets of variables can reflect both theoverall correlation using this correlation to identify characterized in that the compositioncan used for the classification.The vector representation of images is simple, but in the real world, many objects areused to represent the tensor, using vector to represent image loss of the spatial informationof images in a certain extent, singular and when the typical thought analysis often leads tomatrix. So, we put the vector derivation to the matrix, the matrix to the conductivity tensor.All kinds of face recognition algorithm based on tensor because it can keep the spatial andtemporal characteristics of the sample in the feature extraction, so the tensor analysis hasbecome a hot topic.In this paper, from two-dimensional canonical correlation analysis and canonicalcorrelation analysis of analogy, derive the canonical correlation analysis method based ontensor, describes in detail the process of calculation tensor canonical correlation analysis,the algorithm steps are given.Finally, the number of iterations of the algorithm, as well as the recognition rate isdetermined through experiments. Verified based on the fusion tensor CCA near-infrared andvisible face recognition algorithm can improve the success rate of face recognition,especially within the range of the feature extraction, in a small characteristic dimension.