Study On The Extraction Method Of3D Facial Feature

Because of the more consideration of human security, face recognition technology isbecoming the focus of attention of the researchers. Due to the essence of face recognitiontechnology is used facial feature to express face and then used feature to recognize, facialfeature extraction become a key part of the face recognition technology. The common facialfeature extraction mainly based on two dimensional data and three dimensional face data. The3D facial data has more abundant information in the amount than the two-dimensional photo.And illumination, posture, expression and other appendages which strongly impact on thetwo-dimensional image is not obvious to the three-dimensional data. Therefore, more andmore research based on the three-dimensional facial data to do feature extraction.The main content of this article is the research of three-dimensional face featureextraction which based on triangular mesh model. With the research of a variety of featureextraction method, this paper choose Shape Index feature to locate feature points. Then weextracts contour line characteristics and geometry measurement characteristics, and use theEuclidean distance to calculate the similarity to compare the advantage and the disadvantageof these two methods. Finally by compound the two features, our experiments obtain ahigher recognition rate. The main work of this paper is:In the part of "Three-dimensional face registration ", this paper used PCA method forthree-dimensional face rough registration and the candidate point set of the nasal tip location,the method is not only has small amount of computation but also simplify work of the later.In the part of "Three-dimensional face feature point location"，to find the neighborhoodset of feature points, this paper find a new and improved search algorithm.This method useda short time compared to traversal search algorithm and has a higher accuracy compared todirect neighbor search algorithm. To determine candidate points set, we used Shape Indexfeature. Due to the Shape Index feature can characterize bump of the feature pointneighborhood and its value has obvious aggregation, the experiment shows the method isbetter on locate feature points.In the part of "Three-dimensional facial feature extraction ", this paper compoundcontour lines characteristics and geometric measurement characteristics feature. Theexperiments verify the compound feature has a higher recognition rate.