摘要
在基于内容的三维模型检索系统中,特征提取技术是三维模型检索的关键。为此,提出基于局部特征的三维模型检索算法。定义一种新的局部特征描述符:曲度,将其作为三维模型检索时的特征。曲度作为对平均曲率与高斯曲率的校正,在不增加额外计算量的前提下,可同时克服平均曲率对平滑模型的不敏感性和高斯曲率分布较均匀的缺点,更真实地反映三维模型的局部弯曲程度。实验结果表明,以曲度作为特征进行检索,可明显提高检索的查准率,配合全局特征检索时则可在保证查全率的基础上,大幅提高检索的准确性。
Feature extraction is the most important technology in content-based 3D model retrieval. This paper proposes an algorithm for feature extraction based on curvature features of 3D triangle model. The curvature of 3D model can become the sum of Gaussian curvature and mean curvature. So it is a blend of the Gaussian curvature and mean curvature. It can estimate the curvature of 3D model's vertices through the discrete differential geometry,and build a feature vector of 3D model. Experimental results show that the algorithm of this paper is better than the average curvature or Gaussian curvature as the feature vector.
出处
《计算机工程》
CAS
CSCD
北大核心
2015年第3期218-222,共5页
Computer Engineering
基金
国家自然科学基金资助项目(60873094)
关键词
三维模型检索
曲度
高斯曲率
平均曲率
特征提取
3D model retrieval
camber
Gaussian curvature
mean curvature
feature extraction