Research On Three Dimensional Dynamic Face Localization

With the continuous progress of the society,biometric technology has gradually attracted the attention of researchers from all walks of life.Face recognition is one of the hot topics in the field of research because of its naturalness and not easy to be detected by the individual.Because of its huge application prospects in the field of personal information security,self-service,e-commerce,security and defense,Face recognition has been widely applied to many researchers.For decades,researchers have made significant progress in 2D face recognition algorithms.However,almost all of the 2D face recognition methods are extremely sensitive to the following external conditions:lighting,head direction,facial expressions and makeup etc.3D face recognition is to deal with the spatial model of the human face,so it is considered to be safer and more secure.Face localization is the key basic step in face recognition.So the accuracy of the results is very important for the expression extraction and analysis as well as the correction of face.Therefore,the rapid and accurate localization of the face has become a top priority.In this paper,we focus on the spatial model of 3D dynamic human face,and deeply discuss the localization,which is the data processing part of 3d face recognition.It can be divided into three stages,namely,the reading of the face model,the marking of the feature points and the location of the human face.According to the dynamic 3D face model data,we first use VS2010 to read the data and display the face model.This paper first briefly describes the three-dimensional data acquisition equipment: 3D camera.Then,the paper describes the way of reading it into the computer display,which is based on the 3D model.Because the collected face model data by the scanner is large and contains invalid parts such as ears,neck,shoulders etc,so it is necessary to remove the useless part and to cut and locate the active parts of the face that are interested in face recognition.In order to accurately extract the features of faces and reduce the interference of redundant information,the paperselects a sphere to locate the face.In this paper,we first find the feature points in the existing models.The feature extraction of discrimination determines the recognition performance,so feature extraction plays an important role in 3D face recognition.Based on the feature extraction method,three kinds of acquisition methods are given,which are manual location feature points,feature points defined by depth information and classical ASM algorithm.Then the advantages and disadvantages of the three methods are compared,as well as various methods to adapt to the scene.Finally,creating a sphere of its 3d model with the nasal tip point to cut face region,so as to realize the localization of face region.The key point of face acquisition is to calculate the radius of the sphere,the method of geodesic distance is given.Then by comparing it with the most commonly used Euclidean distance,the advantages of geodesic distance can be seen.The experimental results show that the proposed method of 3D face recognition in this paper has good practicability.