Research On The Technology Of Face Recognition With Ageing Variances Based-on SVM

Human face recognition is attractive in pattern recognition and image processing. It can be applied to security system, human ID management, teleconference, digital surveillance and so on. However, the recognition rate will decline sharply due to the shape and texture change in the face for different age. To solve this problem, we considered both age estimation and face image reconstruction, and bring them into face recognition algorithm which based on the present face recognition and age change research. Thereby we present the face recognition which allows age change. We designed and implemented the prototype system of face recognition.The highlights and main contributions of the dissertation include:(1)An age estimation algorithm based on selected Local Binary Pattern and weighted Support Vector Machine regression is proposed. After analyzing the advantages and disadvantages of the existing age estimation method, we extract the key feature from human aging faces using Local Binary Pattern, then use the contribution analysis algorithm of neural network to choose the eigenvector and calculate the weights of the remained feature, at last we use weighted support vector regression method to train and gain the whole age function which established the corresponding relationship between texture feature and the age. Experimental results show that the method can effectively estimate the age of the human faces which the error is within 3 years of age, at the same time provide age measurement for face image reconstruction.(2)A human face image reconstruction for different age based on the shape and texture characteristics of LBP is presented. First, change face shape according to the facial characteristics data, combined with Radial Base Function deformation technology. We get age texture of the elderly face images using the LBP method, and then add it to the changed face image. During the process, we can reconstruct the target age human face by judging the age of human face through the age estimation function, adjusting age texture parameter which can control wrinkle's changes. Experimental results show that the method can livingly reconstruct different ages of human faces.(3)Classification algorithm is studied which based on Gabor wavelet and the Multi-class Support Vector Machines. We utilize Gabor wavelet to extract face features and PCA to reduce the dimensions, and then using Multi-class Support Vector Machines classification algorithm, then came face recognition which allows age changes, and also verified the validity and rationality of face image reconstruction process.(4)A prototype system of face recognition with ageing variances is designed and implemented based on the idea of oriented object. We divide the system into four modules that is image preprocess, age estimation, face image reconstruction and face recognition. And makes the system recognize people according to face image even the training image is collected 5 years earlier than the test image.