电子学报 ›› 2019, Vol. 47 ›› Issue (1): 92-99.DOI: 10.3969/j.issn.0372-2112.2019.01.012

• 学术论文 • 上一篇    下一篇

一种SVM学习框架下的Web3D轻量级模型检索算法

周文, 贾金原   

  1. 同济大学软件学院, 上海 201804
  • 收稿日期:2017-10-09 修回日期:2018-02-15 出版日期:2019-01-25 发布日期:2019-01-25
  • 作者简介:周文 男,1984年2月出生于安徽铜陵,现为同济大学博士研究生,主要研究方向:虚拟现实、三维模型检索等.E-mail:zhouwen606@126.com;贾金原 男,1963年11月出生于山东德州,现为同济大学教授,博士生导师,主要研究方向:虚拟现实、轻量级三维建模等.E-mail:jyjia@tongji.edu.cn
  • 基金资助:
    中央高校基本科研业务费-学科交叉类项目重点类项目A类(No.0200219153)

Web3D Lightweight for Sketch-Based Shape Retrieval Using SVM Learning Algorithm

ZHOU Wen, JIA Jin-yuan   

  1. School of Software Engineering, Tongji University, Shanghai 201804, China
  • Received:2017-10-09 Revised:2018-02-15 Online:2019-01-25 Published:2019-01-25

摘要: 随着Web3D技术的发展,对于互联网检索三维模型的需求越来越迫切,特别是基于草图的模型检索.本文对基于草图的三维模型检索相关技术进行了研究,提出了三维模型轻量化处理算法、基于支持向量机三维模型最佳视点选择算法.本文首先对模型进行简化处理,投影三维模型为多个视点图像.其次,使用支持向量机在草图数据集上学习规则,并根据规则进行相应的视点图像分类,获得最佳视点图像.再次,对视点图像提取梯度直方图特征并进行K-means聚类和索引,减少特征空间,获得三维模型的特征字典.最后,在开源数据集上进行相关的实验并对结果进行分析,相关结果表明方法具有很强鲁棒性、准确性.

关键词: 基于草图的三维模型检索, 支持向量机, 简化, 梯度直方图, 特征字典

Abstract: With the rapid development of Web3D technology,the demand to use and retrieve 3D model is becoming urgent.Especially,sketch-based shape retrieval is very important.In the paper,a framework is proposed,which includes lightweight for shape,SVM-based learning algorithm.In particular,the model is simplified and then projected into multi-views images.Besides,a SVM classifier is used to classify these images.Moreover,histogram of oriented gradient (HOG) features is extracted from the input sketch image.Furthermore,K-means algorithm is used to cluster and index these features in order to generate a features dictionary.Finally,the related experiments are conducted to verify the feasibility of the approach in open source datasets.The result shows that the proposed method is robust and superior,compared with other methods.

Key words: sketch-based shape retrieval, SVM, simplified, HOG, features dictionary

中图分类号: