人脸动态非刚性三维特征提取仿真  

Simulation of Dynamic Non-Rigid 3D Feature Extraction for Human Face

作　　者：谭超[1] 李昕璐[1] TAN Chao;LI Xin-lu(Lushan College,Guangxi University of Science and Technology,Liuzhou Guangxi 545616,China)

机构地区：[1]广西科技大学鹿山学院,广西柳州545616

出　　处：《计算机仿真》2020年第8期389-393,共5页Computer Simulation

基　　金：2020年度广西高校中青年教师科研基础能力提升项目(2020KY60014)。

摘　　要：现有的人脸动态三维特征提取方法存在速度慢、特征识别率差的弊端,为此提出一种人脸动态非刚性三维特征提取方法。首先获取三维人脸图像的点云数据,在获得图像点云坐标的基础上通过归一化过程实现图像预处理;利用两个给定的曲面形状之间的非刚性映射还原人脸形变图像,构建人脸非刚性点集;将采集到的数据利用KL的变换投影方式投射至低维特征空间中,结合成分分析法完成对人脸动态三维特征的提取。仿真结果证明,该方法下的特征提取速度较快,对特征的识别率明显高于传统方法,表明该方法具有明显的应用优势。In this article,a dynamic non-rigid 3D feature extraction method for human face was put forward.Firstly,the point cloud data of 3D face image were obtained,and then the image preprocessing was achieved by normalization.Secondly,the non-rigid mapping between two curved surfaces was used to restore the deformation image of human face and construct the sets of non-rigid points of human face.Thirdly,the collected data were projected into the low-dimensional feature space by KL transformation projection.Finally,dynamic 3D feature extraction of human face was completed by the component analysis method.Simulation results show that the proposed method can extract the features faster than existing methods.Meanwhile,the recognition rate of feature is obviously higher than that of the traditional method.This method has obvious application advantages.

关 键 词：人脸 非刚性三维特征 特征提取 成分分析法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]