%0 Journal Article
%T 基于流形排序的三维模型检索方法<br>A new 3D model retrieval method based on manifold ranking
%A 牟春倩
%A 唐雁
%A 胡金戈&lt
%A br&gt
%A MOU Chunqian
%A TANG Yan
%A HU Jinge
%J 山东大学学报（工学版）
%D 2017
%R 10.6040/j.issn.1672-3961.0.2016.382
%X 摘要： 针对现有的基于视图的三维模型检索方法大多将二维投影视图特征直接用于表示三维模型,忽略不同视点下的二维投影视图特征的贡献度的问题,提出一种基于流形排序的三维模型检索方法,关注不同二维投影视图特征的贡献度。通过旋转三维模型获得34张不同视点的二维投影视图,采用基于尺度不变特征变换(scale invariant feature transform, SIFT)特征的词袋模型提取34张二维投影视图的词频向量特征,利用流形排序将同一个三维模型的34个词频向量特征聚合成一个三维模型特征。试验表明,基于流形排序的三维模型检索方法能够有效地提高检索结果的准确率。<br>Abstract: Most existing view-based 3D model retrieval methods used features of 2D projection views to represent a 3D model directly, ignored their contributions to a 3D model. Therefore, a new 3D model retrieval method based on manifold ranking was proposed, which focused on the contributions of 2D projected views features. 2D projected views from 34 different viewpoints obtained through rotation, word frequency vector featuresextracted by bag-of-feature based on scaleinvariant feature transform(SIFT)features, then aggregated 34 word frequency vector features of a 3D model into a 3D model feature. The experimental results showed that our method improved the retrieval accuracy well
%K 三维模型检索
%K 二维投影视图
%K 流形排序
%K 词袋模型
%K SIFT特征
%K &lt
%K br&gt
%K 2D projected view
%K SIFT feature
%K 3D model retrieval
%K manifold ranking
%K bag-of-feature
%U http://gxbwk.njournal.sdu.edu.cn/CN/10.6040/j.issn.1672-3961.0.2016.382