摘要
为实现已有设计实例的重用,满足对设计基元层次相似搜索的要求,提出了一种三维模型检索方法.在模型B-Rep数据结构基础上,通过构造离散曲面获取三维模型的离散点集.利用正态分布种子作为索引,对离散点集进行随机选取,计算任意两点之间的欧氏距离获得三维模型的欧氏距离序列.通过分治快速排序和概率分布统计获得三维模型的形态分布图,将其作为三维模型的检索索引,并进行了鲁棒性分析.使用BP神经网络对不同模型的形态分布图进行非线性映射,通过比较三维模型的形态分布图获得其形状相似度,实现了设计基元的检索.开发了原型系统,以机械零件的不同粒度设计基元为重点进行验证和应用,实验结果表明,该方法具有较好的可计算性和较高的检测精度,对于实现已有设计重用及提高产品设计效率具有重要意义.
A shape retrieval method of 3D models was put forward to realize the reuse of existing design cases and satisfy the need of hierarchy similarity searching of design primitives. On the basis of the B-Rep data structure of 3D models, the discrete points set is obtained through constructing the discrete surfaces of solid models. The elements of the points set are selected at random via seeds which belong to normal distribution. The Euclidean distance sequence is acquainted by calculating the Euclidean distance between arbitrary two points from the selected points. The shape distribution graph （SDG） is then obtained via the quick sorting based on the divide-and-conquer algorithm and the distribution possibility statistics of the Euclidean distance sequence, then the SDG is defined as the shape retrieval index after robustness analysis. The nonlinear mapping of the SDG of 3D models is achieved by means of BP neural network. The shape similarity between solid models is obtained via the comparison between the SDGs, and the retrieval of design primitives is realized. A prototype system was developed to verify the proposed method focusing on the design primitives with different granularities of mechanical parts. The test results proved that the computability and precision of the method are favorable enough for practical requirements. The method has important significance to realize the design reuse and improve the design efficiency.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2009年第5期877-883,共7页
Journal of Zhejiang University：Engineering Science
基金
国家"863"高技术研究发展计划资助项目(2006AA04Z114)
国家自然科学基金资助项目(50775201)
