基于参数化表示的三维形体分析及其应用研究

项目名称： 基于参数化表示的三维形体分析及其应用研究

项目编号： No.60873137

项目类型： 面上项目

立项/批准年度： 2009

项目学科： 金属学与金属工艺

项目作者： 严京旗

作者单位： 上海交通大学

项目金额： 27万元

中文摘要： 随着三维获取设备的革新以及计算机处理性能的不断提高，三维形体已成为继声音、图像和视频之后的又一种多媒体数据类型，吸引了众多研究者的关注。其中，三维形体分析正成为计算机图形学和数字几何处理方面的热点研究方向之一，在三维形体分类、检索和识别等方面具有重要意义。 本项目开展了几何滤波平滑、几何特征点定位、三维形体匹配及识别等内容研究，提出了基于隐函数拟合的三维形体显著特征点提取，并应用于三维人脸显著特征点定位以及姿态归一化；提出了基于法向最近点的三维形体相似性度量及匹配算法，并应用于多姿态多表情的三维人脸识别；提出了三维掌纹全局形体特征描述以及基于主线配准的三维掌纹识别方法；提出了基于双边滤波的三维形体主曲率及主方向场平滑，并应用于三维脑沟盆地分割和脑沟基底线提取。 在项目执行期间，项目组成员共发表了15篇相关论文，其中SCI收录5篇次，EI收录11篇次。

中文关键词： 三维形体分析;统计学习;三维生物特征识别;几何滤波;几何特征提取

英文摘要： With the innovation of 3D acquisition devices and the development of computer processing capability, 3D shape has become another important media after voice, image and video. A lot of researchers in the computer community have been attracted to study on 3D shape analysis, which is becoming a hot direction in the fields of computer graphics and digital geometry processing. 3D shape analysis is very important for 3D shape classification, retrieval, and recognition. This project has studied geometrical mesh smoothing, geometrical feature point locating, 3D shape matching and recognition. A new method, based on multilevel implicit function approximation, has been proposed for extracting salient feature points on 3D shapes, and has been used for locating salient feature points on 3D faces and normalizing them under varied poses; a shape-similarity measurement based on closest points along their normals has been developed for matching 3D faces under varied poses and varied expression; a global shape description and principal line matching have been argued for 3D palmprint recognition; and the bilateral filter has been developed for smoothing principal curvatures and their corresponding direction flows, which are used for sulcal basins parcellation and sulcal ravines extraction. During this project, the project members have published 15 papers, including 5 SCI-indexed and 11 EI indexed.

英文关键词： 3D Shape Analysis; Statistical Learning; 3D Biometrics; Geometrical Filters; Geometrical Feature Extraction