Page 358 - 《软件学报》2025年第12期

软件学报 ISSN 1000-9825, CODEN RUXUEW E-mail: jos@iscas.ac.cn
2025,36(12):5739−5754 [doi: 10.13328/j.cnki.jos.007416] [CSTR: 32375.14.jos.007416] http://www.jos.org.cn
©中国科学院软件研究所版权所有. Tel: +86-10-62562563

*
局部-全局动态图学习与互补融合的点云配准方法

邱巧燕 1 , 叶海良 1 , 曹飞龙 1 , 吕 科 2

1
(中国计量大学 理学院, 浙江 杭州 310018)
2
(中国科学院大学 工程科学学院, 北京 100049)
通信作者: 曹飞龙, E-mail: caofeilong88@zjnu.edu.cn

摘 要: 现有基于深度学习的点云配准方法主要聚焦于特征提取和特征匹配方面的研究, 然而, 其在特征提取阶段
对局部和全局图结构的挖掘尚不充分, 同时在匹配过程中对差异信息的探索也较为有限. 为此, 提出了一种局部-全
局动态图学习与互补融合的点云配准方法. 具体而言, 动态偏移的局部图学习模块通过构造包含几何和语义信息
的代理点来刻画特征空间中潜在的图结构, 从而获得更具判别性的局部特征. 其次, 设计了动态关注的全局图学习
模块, 根据点之间的相互关系自适应地调整关注权重, 有效地捕获了点云中的长程依赖关系. 为了进一步提高两个
点云之间的对应关系, 构造了注意力驱动的互补融合模块, 根据交叉注意力机制来挖掘相似信息和差异信息, 并利
用自注意力机制优化特征之间的关联性. 实验结果表明, 该方法在公开数据集上实现了最优的配准效果, 并具备良
好的计算效率.
关键词: 深度学习; 点云配准; 图神经网络; 特征提取; 特征匹配
中图法分类号: TP181

中文引用格式: 邱巧燕, 叶海良, 曹飞龙, 吕科. 局部-全局动态图学习与互补融合的点云配准方法. 软件学报, 2025, 36(12):
5739–5754. http://www.jos.org.cn/1000-9825/7416.htm
英文引用格式: Qiu QY, Ye HL, Cao FL, Lyu K. Point Cloud Registration Method Based on Local-global Dynamic Graph Learning
and Complementary Fusion. Ruan Jian Xue Bao/Journal of Software, 2025, 36(12): 5739–5754 (in Chinese). http://www.jos.org.cn/
1000-9825/7416.htm

Point Cloud Registration Method Based on Local-global Dynamic Graph Learning and
Complementary Fusion
1
1
1
QIU Qiao-Yan , YE Hai-Liang , CAO Fei-Long , LYU Ke 2
1
(College of Sciences, China Jiliang University, Hangzhou 310018, China)
2
(School of Engineering Science, University of Chinese Academy of Sciences, Beijing 100049, China)
Abstract: Existing deep learning-based point cloud registration methods primarily focus on feature extraction and feature matching.
However, the exploration of local and global graph structures during the feature extraction stage remains insufficient, and the investigation
of difference information during the matching process is also limited. To address these issues, this study proposes a point cloud
registration method based on local-global dynamic graph learning and complementary fusion. Specifically, the dynamic offset-based local
graph learning module characterizes the underlying graph structure in the feature space by constructing proxy points that contain both
geometric and semantic information, leading to more discriminative local features. In addition, a dynamic attention-based global graph
learning module is designed, which adaptively adjusts attention weights based on the relationships between points, effectively capturing
long-range dependencies in the point cloud. To further enhance the correspondence between the two point clouds, the attention-driven
complementary fusion module utilizes the cross-attention mechanism to extract similar and distinctive information, while applying the self-
attention mechanism to refine the relationships between features. Experimental results demonstrate that the proposed method achieves
optimal registration performance on public datasets while maintaining acceptable computational efficiency.

* 基金项目: 国家自然科学基金 (62032022, 62176244)
收稿时间: 2024-12-13; 修改时间: 2025-01-10; 采用时间: 2025-02-18; jos 在线出版时间: 2025-09-03
CNKI 网络首发时间: 2025-09-04