基于谱变换的非刚性模型局部匹配  

Partial Matching of Non-rigid Shape based on Spectral Transformation

作　　者：韩丽[1] 徐圣斯 朴京钰 HAN Li;XU Sheng-si;PIAO Jing-yu(School of Computer and Information Technology,Liaoning Normal University,Dalian Liaoning 116081,China)

机构地区：[1]辽宁师范大学计算机与信息技术学院,辽宁大连116081

出　　处：《软件》2020年第10期22-26,共5页Software

基　　金：国家自然科学基金资助项目(61702246);辽宁省社会科学基金资助项目(2018lslktyb-084);大连市科技创新基金项目(2019J12GX038)。

摘　　要：面向海量复杂的非刚性三维模型的匹配与检索需求,提出了基于谱变换的局部匹配方法,有效提高了局部检索的性能,为进一步实现三维模型的形状编辑、修补以及创新合成奠定了坚实的基础。首先,基于离散拉普拉斯映射方法,通过ICP配准方法有效分析二维谱图,提取最优的谱特征表示;其次,构建谱特征的形状描述算子,计算谱空间的几何变换参数;进而,引入区域查询实现基于谱变换的区域配准;最终,通过区域搜索与相似性度量实现高效的非刚性模型局部匹配。实验结果表明,本方法不仅能够有效识别同一类非刚性模型的局部结构,而且对于不完整的模型匹配具有较强的鲁棒性,具有更优的局部检索性能。Aiming at the requirements of shape matching and shape retrieval of massive non-rigid 3D shapes,a novel partial matching method based on spectral transformation was proposed,it effectively improved the performance of shape retrieval and provided a solid foundation for further shape editing,mending and innovative synthesis of 3D models.First,the optimal two-dimensional spectral feature representation of non-rigid 3D shape is extracted based on the spectrum theory.Secondly,we construct shape context description of the spectral features and calculate the geometric transformation parameters of two spectral graphs;Furthermore,a region query is introduced to implement local region registration based on spectral transformation;Finally,partial matching of non-rigid shapes is achieved through local region searching and similarity estimation.Experimental results show that our method not only effectively improves the efficiency of model matching,but also can effectively identify the structural features of the same type of shape,and at the same time has strong robustness for matching between incomplete shapes.

关 键 词：谱变换 局部匹配 非刚性模型 区域配准 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]