Three-dimensional Face Recognition Based On Information Fusion

Of all the biometric features, due to face recognition have such advantages as most natural, friendly, less interference and user acceptance that it remains one of the most active research topics in pattern recognition. Automatic face recognition technology is the technology that computer analyse face images ,extract effective information and make a decision with a classifier.In the past several decades, most work focuses on the source of 2D intensity or color images. Since the accuracy of 2D face recognition is influenced by variations of poses, expressions, illuminations and subordinates, it is still difficult to develop a robust automatic 2D face recognition system. The 3D facial data can provide a promising way to understand the characteristics of the human face in 3D domain, and has potential possibility to improve the performance of the recognition system.There are three parts in face recognition including the preprocessing of face images, feature extraction and the design of classifier. The main work in this article are as following:Firstly, the acquisition of dual-mode data.Detect the face area from the whole region including neck,shoulders,hair and other background. Finish the work of positioning the nose tip,eliminating the noise , filling the holes and ICP registration.To reduce the influnce of expression ,we use the depth image after selecting the robust area as normalized 3D image.Secondly, research on Support Vector Machines (SVM) and RBF Neural Network Classifier. compare the recognition rate on 2D and 3D data using SVM, and use RBF neural network classifier on 3D face classification for the first time.Thirdly, dual-mode information fusion algorithms research. Compare the recognition performance between 2D and 3D data under different light conditions and expressions, propose information fusion algorithm to improve recognition performance.