模式识别与人工智能
2025年4月4日 星期五   首 页     期刊简介     编委会     投稿指南     伦理声明     联系我们                                                                English
模式识别与人工智能  2022, Vol. 35 Issue (3): 271-282    DOI: 10.16451/j.cnki.issn1003-6059.202203007
研究与应用 最新目录| 下期目录| 过刊浏览| 高级检索 |
面向三维目标的矢量型卷积网络
邱起璐1, 赵杰煜1,2, 陈瑜1
1.宁波大学 信息科学与工程学院 宁波 315211;
2.浙江省移动网应用技术重点实验室 宁波 315211
A Convolutional Vector Network for 3D Mesh Object Recognition
QIU Qilu1, ZHAO Jieyu1,2, CHEN Yu1
1. Faculty of Electrical Engineering and Computer Science, Ning-bo University, Ningbo 315211;
2. Mobile Network Application Technology Key Laboratory of Zhe-jiang Province, Ningbo 315211

全文: PDF (962 KB)   HTML (1 KB) 
输出: BibTeX | EndNote (RIS)      
摘要 目前,三维目标识别方法大多基于卷积神经网络,在特征聚合过程中过多使用池化层,导致三维目标的空间信息丢失.针对上述问题,文中提出面向三维目标的矢量型卷积网络,用于完成三维目标的识别.首先,使用曲面多项式拟合网格目标的局部区域.然后,使用聚类算法得出曲面形状卷积核,通过卷积核和目标表面的相似度度量生成结构感知的特征向量,再利用多头自注意力机制模块实现局部区域到更大范围的特征聚集,得到部件层次特征向量.最后,使用三维矢量型网络实现目标分类.文中网络在SHREC10、SHREC11、SHREC15数据集上均取得较高的分类精度.此外,多分辨率目标对比实验和多采样点对比实验验证文中网络具有较强的泛化性和鲁棒性.
服务
把本文推荐给朋友
加入我的书架
加入引用管理器
E-mail Alert
RSS
作者相关文章
邱起璐
赵杰煜
陈瑜
关键词 三维形状识别矢量型卷积网络结构感知三维矢量网络    
Abstract:The 3D object recognition is usually based on convolutional neural networks. The spatial information of 3D objects is lost due to too many pooling layers used in the process of feature aggregation. To solve the above problem, a convolutional vector network for 3D mesh object recognition is proposed in this paper. Firstly, the surface polynomials are introduced to fit the local mesh of the object, and then the surface shape convolution kernels are clustered. The feature vector of structure-aware is generated by self-measuring the similarity between the local mesh in the object and convolution kernels. The multi-headed attention mechanism module is then employed to achieve feature aggregation from local regions to a larger scale to obtain component-level feature vectors. Finally, the 3D objects are classified through the 3D vector network. The proposed network achieves a high classification accuracy on SHREC10, SHREC11 and SHREC15 datasets. In addition, generalization and robustness of the proposed network are demonstrated through the multi-resolution object comparison experiment and the multi-sampling point comparison experiment.
Key words3D Shape Recognition    Vector-Type Convolutional Network    Structure-Aware    3D Vector Network   
收稿日期: 2021-10-29     
ZTFLH: TP 391.4  
  TP 181  
基金资助:国家自然科学基金项目(No.62071260,62006131,61603202)、浙江省自然科学基金项目(No.LZ22F020001,LY17F030002,LY20F030005)资助
通讯作者: 赵杰煜,博士,教授,主要研究方向为计算智能、模式识别、人机自然交互.E-mail: zhao_jieyu@nbu.edu.cn.   
作者简介: 邱起璐,硕士研究生,主要研究方向为三维卷积网络、模式识别.E-mail:jdqql@163.com.
陈 瑜,博士研究生,主要研究方向为图像处理、深度学习.E-mail:chenyu_cycy@126.com.
引用本文:   
邱起璐, 赵杰煜, 陈瑜. 面向三维目标的矢量型卷积网络[J]. 模式识别与人工智能, 2022, 35(3): 271-282. QIU Qilu, ZHAO Jieyu, CHEN Yu. A Convolutional Vector Network for 3D Mesh Object Recognition. Pattern Recognition and Artificial Intelligence, 2022, 35(3): 271-282.
链接本文:  
http://manu46.magtech.com.cn/Jweb_prai/CN/10.16451/j.cnki.issn1003-6059.202203007      或     http://manu46.magtech.com.cn/Jweb_prai/CN/Y2022/V35/I3/271
版权所有 © 《模式识别与人工智能》编辑部
地址:安微省合肥市蜀山湖路350号 电话:0551-65591176 传真:0551-65591176 Email:bjb@iim.ac.cn
本系统由北京玛格泰克科技发展有限公司设计开发 技术支持:support@magtech.com.cn