基于深度学习的点云语义分割研究综述

A Survey of Semantic Segmentation of Point Cloud Based on Deep Learning

点云作为一种重要的三维数据类型,在自动驾驶、机器人、虚拟及增强现实等人工智能方面应用广泛。点云语义分割是点云处理中的关键任务,旨在将点云中的每个点分配给特定的语义类别。综述了国内外基于深度学习的点云语义分割的研究进展。首先,总结了点云语义分割中常用的开源数据集,并介绍了间接基于点云和直接基于点云的深度学习处理方法及其应用进展。此外,给出了这些方法的实验结果,并对他们进行了简要对比。最后,对当前点云语义分割所存在的问题进行了探讨,并提出了未来的研究发展方向。

PointCloud is an important type of 3D data that is widely used in artificial intelligence applications such as autonomous driving,robotics,virtual and augmented reality.Point cloud semantic segmentation is a key task in point cloud processing,which aims to assign each point in the point cloud to a specific semantic category.This paper reviews the research progress of deep learning-based point cloud semantic segmentation in domestic and foreign studies.Firstly,commonly used open source datasets in point cloud semantic segmentation are summarized,and both indirect and direct deep learning processing methods based on point cloud are introduced along with their application progress.Additionally,the experimental results of these methods are presented and briefly compared.Finally,the problems in current point cloud semantic segmentation are discussed,and future research directions are proposed.