项目名称: 三维信息中形状基元的识别、提取及应用
项目编号: No.61271431
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 无线电电子学、电信技术
项目作者: 张义宽
作者单位: 中国科学院自动化研究所
项目金额: 80万元
中文摘要: 3D建模的高效性和高精细性是多媒体信息处理、虚拟现实、逆向工程等研究中的一个亟待解决的技术挑战。本项目根据拓扑知觉理论,按照视觉认知自顶向下的整体性,利用机器学习分割与标注3D模型中的实体对象,采用RANSAC、流形学习等方法识别与提取3D形状基元,通过高维数据集聚类等提取相似性规则和基本相似单元,建立基于形状基元或基本相似单元的层级结构,实现3D对象或场景三维信息的与图形认知层次一致的理解和高效表达。有利于实现相应三维信息的高效重建、渲染、编辑等。主要研究形状基元和相似性规则的准确快速提取、基于形状基元的层级结构分析、基于层级结构的3D重建等。主要创新是快速、准确的形状基元提取方法;3D模型相似性规则的高效提取方法;基于形状基元、相似性规则及相似基本单元的3D模型层次结构的构建等。技术难点是如何准确快速地提取形状基元和相似性规则、构建符合视知觉信息感知组织和分类融合基本原理的层级结构。
中文关键词: 优化采样;形状基元识别与提取;相似性规则;层级结构;布局空间
英文摘要: Efficiency and high precision in 3D modeling is an important challenge in multimedia information processing, virtual reality, reverse engineering, and others. The basic idea of this project is to segment and label the entity objects in 3D model by machine learning techniques. The 3D shape primitives are detected and extracted by using RANSAC and manifold learning, and the similarity regularities and basically similar units are extracted through clustering of the 3D data set. Then the hierarchical structure is created based on shape primitives or basically similar units, according to the topological perception theory and the top-down integrity of visual cognition in this project. The hierarchical structure is adopted as an understanding and an efficient expression of the 3D information of objects or scenes, and this understanding and expression are consistent to the graphical cognitive level. Such a hierarchical structure is also conducive to efficient reconstruction, rendering, and editing of the corresponding 3D shape information. The main research contents of this project are: quick and accurate extraction of shape primitives and similarity regularities, hierarchical structure analysis based on the shape primitives, and 3D reconstruction based on the hierarchical structure. The main innovations are rapid and a
英文关键词: Optimal sampling;Recognition and extraction of shape primitive;Similarity regularities;Hierarchical structure;Layout space