应用于无人车的三维点云分类方法研究综述  

A Survey of Three-Dimensional Point Cloud Classification Methods Applied to Unmanned Vehicles

作　　者：周梦飞 王子洋 杨家富[1] ZHOU Mengfei;WANG Ziyang;YANG Jiafu(Mechanical and Electronic Engineering,Nanjing Forestry University,Nanjing Jiangsu 210037,China)

机构地区：[1]南京林业大学机械电子工程,江苏南京210037

出　　处：《传感技术学报》2022年第7期931-937,共7页Chinese Journal of Sensors and Actuators

基　　金：江苏省研究生科研与实践创新计划项目(SJCX22_0314)。

摘　　要：三维点云是无人车环境感知任务中激光雷达采集到的重要数据,包含了周围环境和物体的属性特征。点云分类作为环境感知任务中关键一环,目的是根据属性特征对周围物体和环境进行种类标定,分类结果的准确与否关系到后续导航定位和路径规划任务。在点云分类任务分析研究基础上,依据国内外研究对分类方法进行总结归纳,分别阐述了三种分类方法的思路,详细介绍了三种方法中常用的分类方式及其改进研究,并对其进行优缺点分析总结,根据现有分类方法的研究状况,总结了点云分类方法的研究重点和未来发展趋势。Three dimensional point cloud contains the important data collected by LiDAR in unmanned vehicle environment sensing task,which includes the attribute characteristics of surrounding environment and objects.As a key part of the environment perception task,point cloud classification aims to calibrate the types of surrounding objects and environment according to the attribute characteristics.The accuracy of classification is related to the subsequent navigation,positioning and path planning tasks.Based on the analysis and research of point cloud classification tasks,the classification methods are summarized and classified according to the research at home and abroad.The ideas of the three classification methods are expounded,the common classification methods and their improvement research are introduced in detail,their advantages and disadvantages are analyzed and summarized,and the research focus and future development trend of point cloud classification methods are summarized according to the research of existing classification methods.

关 键 词：环境感知 激光雷达 点云分类 分类器 深度学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]