基于特征描述子推断能力的三维CAD模型检索算法  

Feature Combination for 3D CAD Model Retrieval Based on Inference Ability of Feature Descriptors

作　　者：石民[1] 张树生[1] 李亮[1] 白晓亮[1] 

机构地区：[1]西北工业大学现代设计与集成制造技术教育部重点实验室,西安710072

出　　处：《机械科学与技术》2012年第11期1864-1867,共4页Mechanical Science and Technology for Aerospace Engineering

基　　金：国家“863”计划项目(2007AA04Z137)资助

摘　　要：针对单一形状特征对三维CAD模型描述能力不足的问题,提出了一种基于特征描述子推断能力进行多特征融合的三维CAD模型检索算法。首先,对三维模型使用不同的特征提取算法提取出相应的特征描述子;然后利用模型库中已有的模型分类信息对这些特征描述子的推断能力进行评估;根据评估结果,建立一个基于贝叶斯条件概率的特征融合框架进行特征融合;最后通过比较融合后所得到的模型多特征描述子,获得模型间的相似性评价,实现对模型的相似性检索。实验结果表明:本文算法对三维CAD模型的检索效果优于使用单一形状特征的检索算法和基于权值进行特征融合的检索算法。A new feature combination algorithm for 3D CAD model retrieval based on the inference ability of shape feature descriptors was proposed. Initially, different feature extraction algorithms were employed to extract the cor- responding feature descriptors of 3D models. Next, the inference ability of those descriptors was evaluated by using the pre-classified information in model database. Then, a feature combination frame based on Bayesian conditional probability was established for multi-feature combination. The similarity of two models was measured by the likeness of their corresponding multi-feature descriptors obtained from the feature combination. The experimental results show the effectiveness of the presented method.

关 键 词：基于内容的三维检索 特征融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]