摘要
随着三维扫描技术的逐渐成熟,三维人体扫描模型的骨骼提取逐渐成为虚拟人建模研究领域的热点之一.现有的三维人体模型骨骼提取方法,存在手工标注任务繁重、对模型姿态过于敏感、计算结果不准确等问题.提出了一种新的三维人体模型的骨骼提取算法:首先,根据Morse原理,将测地距离作为Morse函数的要素,实现姿态无关的人体模型特征点以及拓扑结构的提取;其次,将测地距离等值面作为基础数据,采用截面似圆性判别准则提取模型关节中心所在等值面,从而获得关节中心的准确结果.实验结果表明,与已有算法相比,该方法具有模型姿态无关、计算结果准确等特性,并且能够完全自动地提取三维人体扫描模型的骨骼.
With the development of 3D scanning technique,joint extraction from scanned human body model is one of the most active research areas in virtual human modeling.Many different approaches have been proposed with the aim of extracting joints from scanned human body.But most of these methods can not guarantee pose-independence,while other methods which can ensure pose-independence require manual intervention.To solve this problem,a new framework is presented for automatic pose-independent joint extraction from scanned human body.Firstly,a new Morse function is defined on the scanned human body shape as geodesic distance from points to a source point which can be extracted by a heuristic method.Secondly,the Morse function is calculated,and then feature points and topological structure of the model shape can be extracted automatically according to Morse theory.Finally,shape is divided into segments based on Morse function isolines,and joints can be extracted from the corresponding segments of the human body shape by analyzing the circularity of Morse function isolines of the model.Experiments have been done on 20 scanned human bodies with about 5000 faces and 2300 points in each body shape.The experiment results demonstrate that this method is a pose-independent and automatic method,and is more accurate than previous methods.
出处
《计算机研究与发展》
EI
CSCD
北大核心
2008年第7期1249-1258,共10页
Journal of Computer Research and Development
基金
国家自然科学基金项目(60403042,60573162,60703019)
国家“九七三”重点基础研究发展规划基金项目(2002CB312104)
