计算机与数字工程2025，Vol.53Issue(4)：1038-1043,6.DOI:10.3969/j.issn.1672-9722.2025.04.021
基于局部邻域特征自适应的三维点云分类方法研究
A 3D Point Cloud Classification Method Based on Adaptive Local Neighborhood Features
摘要
Abstract
Due to the difference of point cloud density in different areas and the fact that the point cloud data may become sparse after sampling,the neighborhood relationship of the original target point cloud data is destroyed to some extent,thus affect-ing the full expression of the local structural features of the target and the ability to learn the features.To address this problem,a 3D point cloud recognition method based on adaptive optimization of local neighborhood features is proposed,which is suitable for flexi-bly updating the local geometric structure,enhancing the perceptual capability of the network,and better learning 3D point cloud features.First,the uniqueness of each individual point is increased by considering their respective optimal 3D neighborhoods.Sec-ondly,a neighborhood aggregation module based on the construction graph is used to extract local structural features for recognition in an end-to-end manner.Experimental results on the ModelNet40 dataset demonstrate the effectiveness of the method in typical 3D object shape classification.关键词
三维点云/邻域自适应/动态图卷积/特征提取Key words
3D point cloud/neighborhood adaption/dynamic graph convolution/feature extraction分类
计算机与自动化引用本文复制引用
宋振杰,祁云嵩..基于局部邻域特征自适应的三维点云分类方法研究[J].计算机与数字工程,2025,53(4):1038-1043,6.基金项目
国家自然科学基金项目(编号:61471182)资助. （编号:61471182）
