摘要
三维模型坐标归一化是三维模型检索中的一个重要预处理步骤,其重点是旋转归一化.提出一种基于模型表面的法线分布特性来调整连续主成分分析方法选取的旋转主轴的方法.从人的认知习惯出发,将三维模型的表面法线分布特性归为无明显的法线分布方向、具有显著的最大法线分布方向以及具有2个显著的法线分布方向3类;借助法线分布直方图对模型的法线分布特性进行分析和归类,并根据归类结果采用有针对性的主轴修正策略.实验结果表明,与连续主成分分析方法相比,文中方法能得到更理想的旋转归一化结果.
The pose adjustment of 3D models is an important preprocess for 3D model retrieval, while its focus is on rotation normalization. An approach to further adjust the rotation axes determined by Continuous PCA is presented according to the surface normal distributions. Based on people's cognition convention, the normal distributions of 3D models are classified into three cases: with no obvious normal clustering, with one obvious direction of normal clustering, with two obvious directions of normal clustering. We then make the classification based on the normal distribution histograms and adjust the rotation axes conforming to the classification result. Experimental results show that our approach can produce more reasonable rotation normalization results than continuous PCA.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2008年第6期683-688,共6页
Journal of Computer-Aided Design & Computer Graphics
基金
国家"九七三"重点基础研究发展规划项目(2006CB303105)
国家自然科学基金(60602030)
北京交通大学科技基金资助项目(2007RC049)
关键词
连续主成分分析
主轴
法向
旋转归一化
三维模型检索
continuous principal component analysis principal axis
normal vector
rotationnormalization
3D model retrieval