点云分类方法综述  

A Survey of Point Cloud Classification

作　　者：利满雯 赵艳明[1] 李绍彬[1] 冯燕 LI Manwen;ZHAO Yanming;LI Shaobin;FENG Yan(School of Information and Communication Engineering,Communication University of China,Beijing 100024,China)

机构地区：[1]中国传媒大学信息与通信工程学院,北京100024

出　　处：《电视技术》2022年第5期1-8,共8页Video Engineering

基　　金：国家重点研发计划项目“混合现实环境下的观演空间支撑技术”(No.2018YFB1404100)。

摘　　要：点云作为一种常见的三维数据,应用广泛、研究前景巨大。点云分类更是当前的研究热点之一。本文将现有的点云分类方法归纳为传统的基于人工标记点云提取特征描述符和基于深度学习实现点云分类两大类,着重介绍基于深度学习实现点云分类的经典、前沿算法,总结各类算法的基本思想和优缺点,最后分析了目前主要的实现结果并提出点云分类的前景和展望。As a kind of common three-dimensional data,point cloud is widely used and has great research prospects.Point cloud classification is one of the research hotspots.This paper classifies the existing point cloud classification methods into two categories:the traditional extraction of feature descriptors based on artificially labeled point clouds and point cloud classification methods based on deep learning.It focuses on the classic and state of the art algorithms for point cloud classification based on deep learning.It summarizes the basic ideas,advantages and disadvantages of various algorithms,and finally analyzes the main implementation results and puts forward the prospects.

关 键 词：点云分类 深度学习 特征提取 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]