Journal of Computer Applications ›› 2025, Vol. 45 ›› Issue (9): 3003-3010.DOI: 10.11772/j.issn.1001-9081.2024091254

• Multimedia computing and computer simulation • Previous Articles    

Point cloud classification and segmentation network based on dual attention mechanism and multi-scale fusion

Weigang LI1,2, Jiale SHAO1(), Zhiqiang TIAN2   

  1. 1.School of Information Science and Engineering,Wuhan University of Science and Technology,Wuhan Hubei 430081,China
    2.Engineering Research Center for Metallurgical Automation and Measurement Technology of Ministry of Education,Wuhan University of Science and Technology,Wuhan Hubei 430081,China
  • Received:2024-09-05 Revised:2024-10-16 Accepted:2024-10-18 Online:2024-10-31 Published:2025-09-10
  • Contact: Jiale SHAO
  • About author:LI Weigang, born in 1977, Ph. D., professor. His research interests include industrial process control, artificial intelligence, machine learning.
    TIAN Zhiqiang, born in 1996, Ph. D. candidate. His research interests include computer vision.
  • Supported by:
    Hubei Provincial Science and Technology Talent Serving Enterprise Project(202400288)

基于双注意力机制和多尺度融合的点云分类与分割网络

李维刚1,2, 邵佳乐1(), 田志强2   

  1. 1.武汉科技大学 信息科学与工程学院,武汉 430081
    2.武汉科技大学 冶金自动化与检测技术教育部工程研究中心,武汉 430081
  • 通讯作者: 邵佳乐
  • 作者简介:李维刚(1977—),男,湖北咸宁人,教授,博士,主要研究方向:工业过程控制、人工智能、机器学习; 深度学习、点云数据处理
    田志强(1996—),男,湖北武汉人,博士研究生,主要研究方向:计算机视觉。
  • 基金资助:
    湖北省科技人才服务企业项目(202400288)

Abstract:

The existing networks are difficult to learn local geometric shape information of point clouds effectively, and have problems such as being unable to focus on important feature structure effectively and insufficient fusion. Therefore, a point cloud classification and segmentation network based on Dual Attention Mechanism (DAM) and multi-scale fusion was proposed. Firstly, in the data feature extraction stage, geometric positions and weights of the convolution kernels were adjusted using Geometric Adaptive Convolution (GAC) dynamically, so that it was able to adapt to local geometric structure of the point cloud data dynamically, thereby capturing local features more effectively. Secondly, in order to further improve the feature expression ability, the DAM was introduced to learn and adjust weights of the feature channels and spatial information automatically, thereby enhancing feature representation of the key points. Finally, feature information of different scales was connected for effective fusion to enhance the feature learning effect, thereby making the final feature representation richer and improving classification and segmentation accuracy of the network. Experimental results on ModelNet40, ShapeNet and S3DIS datasets show that the proposed network increases the Overall Accuracy (OA) and mean Intersection over Union (mIoU) compared with PointNet++ and DGCNN (Dynamic Graph Convolutional Neural Network), improving the performance of point cloud classification and segmentation effectively.

Key words: point cloud, classification and segmentation, deep learning, attention mechanism, feature fusion

摘要:

现有的网络难以有效学习点云局部的几何形状信息,存在无法有效关注重要特征结构和融合不充分等问题。因此,提出一种基于双注意力机制(DAM)和多尺度融合的点云分类与分割网络。首先,在数据特征提取阶段利用几何自适应卷积(GAC)动态地调整卷积核的几何位置和权重,使它能够动态适应点云数据的局部几何结构,从而更有效地捕捉局部特征;其次,为了进一步提升特征表达能力,引入DAM自动学习并调整特征通道和空间信息的权重,从而增强关键点的特征表示;最后,连接不同尺度的特征信息以进行有效融合,从而增强特征学习效果,使得最终的特征表示更加丰富,以提高网络的分类分割精度。在ModelNet40、ShapeNet和S3DIS数据集上的实验结果表明,所提网络与PointNet++和DGCNN(Dynamic Graph Convolutional Neural Network)相比,总体分类精度(OA)和平均交并比(mIoU)更好,有效提升了点云分类与分割的性能。

关键词: 点云, 分类分割, 深度学习, 注意力机制, 特征融合

CLC Number: