《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (6): 1818-1825.DOI: 10.11772/j.issn.1001-9081.2022050688

• 人工智能 • 上一篇    下一篇

基于特征增强的三维点云语义分割

鲁斌1,2, 柳杰林1,2()   

  1. 1.华北电力大学 计算机系,河北 保定 071003
    2.复杂能源系统智能计算教育部工程研究中心(华北电力大学),河北 保定 071003
  • 收稿日期:2022-05-13 修回日期:2022-08-05 接受日期:2022-08-08 发布日期:2023-06-08 出版日期:2023-06-10
  • 通讯作者: 柳杰林
  • 作者简介:鲁斌(1975—),男,宁夏银川人,教授,博士,CCF高级会员,主要研究方向:智能计算、计算机视觉、综合能源系统
    柳杰林(1997—),男,河北沧州人,硕士研究生,主要研究方向:智能计算、计算机视觉Email:jielin_liu2022@163.com

Semantic segmentation for 3D point clouds based on feature enhancement

Bin LU1,2, Jielin LIU1,2()   

  1. 1.Department of Computer,North China Electric Power University,Baoding Hebei 071003,China
    2.Engineering Research Center of Intelligent Computing for Complex Energy Systems,Ministry of Education (North China Electric Power University),Baoding Hebei 071003,China
  • Received:2022-05-13 Revised:2022-08-05 Accepted:2022-08-08 Online:2023-06-08 Published:2023-06-10
  • Contact: Jielin LIU
  • About author:LU Bin, born in 1975, Ph. D., professor. His research interests include intelligent computing, computer vision, integrated energy systems.

摘要:

为挖掘感知点云几何特征并通过特征增强的方式进一步提高点云语义分割效果,提出了一种基于特征增强的点云语义分割网络。首先,通过设计点云的几何特征感知(GFSOP)模块赋予网络点云局部几何结构的感知能力,捕获点间的空间特征以强化语义表征,并利用分层提取特征思想获得多尺度特征。同时,使用空间注意力和通道注意力融合预测点云语义标签,并通过强化空间关联性和通道依赖性提升分割性能。在室内数据集S3DIS(Stanford large-scale 3D Indoor Spaces)上的实验结果显示,所提网络相较于PointNet++在平均交并比(mIoU)上提升了5.7个百分点,在总体准确度(OA)上提升了3.1个百分点,且在存在噪声、点云密度不均和边界不清晰等问题的点云上表现出更强的泛化性能和更加鲁棒的分割效果。

关键词: 点云, 语义分割, 特征增强, 几何特征, 注意力机制

Abstract:

In order to mine and sense the geometric features of point clouds and further improve the semantic segmentation effect of point clouds by feature enhancement, a point clouds semantic segmentation network based on feature enhancement was proposed. Firstly, the Geometric Feature Sensing Of Point cloud (GFSOP) module was designed to make the network capable of sensing the local geometric structure of point clouds, semantic representations were enhanced by capturing spatial features between points, and multi-scale features were obtained by the idea of hierarchical extraction of features. At the same time, spatial attention and channel attention were fuseed to predict semantic labels of point clouds, and the segmentation performance was improved by strengthening spatial correlation and channel dependence. Experimental results on the indoor dataset S3DIS (Stanford large-scale 3D Indoor Spaces) show that compared with PointNet++, the proposed network improves the mean Intersection over Union (mIoU) by 5.7 percentage points and the Overall Accuracy (OA) by 3.1 percentage points, and has stronger generalization performance and more robust segmentation effect on point clouds with problems of noise, uneven point cloud density and unclear boundaries.

Key words: point cloud, semantic segmentation, feature enhancement, geometric feature, attention mechanism

中图分类号: