基于语义矩阵反馈的多特征融合三维模型检索方法  

3D model retrieval method with multi-feature fusion based on semantic matrix feedback

作　　者：胡敏[1] 罗珣[1] 马韵洁[1] 

机构地区：[1]合肥工业大学计算机与信息学院合肥230009

出　　处：《电子测量与仪器学报》2012年第4期325-330,共6页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(60773043);安徽省自然科学基金(090412059)

摘　　要：为解决相关反馈三维模型检索方法存在用户不能确定模型是否相似的问题,提出了一种基于语义矩阵反馈的多特征融合三维模型检索方法。首先,采用形状分布和球面调和两种特征提取算法进行多特征提取。然后,对每种特征进行检索计算,将得到的相似度进行基于语义的反馈,根据反馈结果对不同特征分配不同的权值。最后,对迭代反馈结果的权求和得到检索模型的相似度。实验结果表明,本方法的检索结果比用单一的特征提取方法得到的结果准确。When the relevance feedback method is used to retrieve the 3D model, it has the problem that the user unclear whether the models are similar or not similar. In order to solve this problem, an integrated method of 3D model retrieval is proposed which based on the combination of semantic matrix feedback and feature. Firstly, 3D model features are extracted by using shape distributions and spherical harmonics methods. Then these features similarity of the 3D mod- els are involved to calculate the assessment weights. These assessment weights are combined with semantic matrix feed- back in the 3D model retrieval. Finally, the similarity of the 3D models is calculated based on the iterative feedback result. The experiment results show that this method is more accurate than the single feature extraction method.

关 键 词：三维模型检索 语义矩阵 相关反馈 特征融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]