Computer Engineering and Applications ›› 2020, Vol. 56 ›› Issue (7): 240-246.DOI: 10.3778/j.issn.1002-8331.1909-0094

Previous Articles     Next Articles

3D Model Retrieval Based on Sphere Projection

PAN Ting, YANG Qiuxiang, JING Zhiyu   

  1. College of Software, North University of China, Taiyuan 030051, China
  • Online:2020-04-01 Published:2020-03-28



  1. 中北大学 软件学院,太原 030051


A 3D model retrieval method based on sphere projection is proposed in order to solve the problems of domain mismatch and how to select views when sketch retrieves 3D models. Firstly, two-dimensional view acquisition method based on sphere projection is proposed according to the domain mismatch problem, and the Gaussian difference and Bezier curve are used to extract the line graph. Secondly, a classifier is constructed by using the relationship between the sketch and the projected image to obtain an optimal view of the model. Finally, the features of the sketch and 2D view are obtained through two Siamese networks, and the joint Bayesian method is used to fuse the output of the sketch and the 2D view to obtain the final result. The experiments prove the feasibility of the method and has better retrieval effect than other methods.

Key words: 3D model, sketch retrieval, projection view, optimal view


针对草图检索三维模型时存在的域不匹配和如何选取视图等问题,提出一种基于球体投影的三维模型检索方法。针对域不匹配问题,提出基于球体投影的二维视图获取方法,并使用高斯差分和贝塞尔曲线完成线图的提取;利用草图和投影图像之间的关系构建分类器,以获取模型的最优视图;通过两个Siamese网络获取草图和二维视图的特征,并用联合贝叶斯(Joint Bayesian)方法来融合二者的输出,从而获得最终结果。实验证明了该方法的可行性,与其他方法相比具有更好的检索效果。

关键词: 三维模型, 草图检索, 投影视图, 最优视图