室内障碍物要素图构建方法  

Construction method of indoor obstacle element map

作　　者：张英南 李景文 姜建武[1,2] ZHANG Yingnan;LI Jingwen;JIANG Jianwu(College of Geomatics and Geoinformation,Guilin University of Technology,Guilin,Guangxi 541004,China;Ecotemporal Big Data Sensing Service Laboratory(Guilin University of Technology),Guilin,Guangxi 541004,China)

机构地区：[1]桂林理工大学测绘地理信息学院,广西桂林541004 [2]生态时空大数据感知服务重点实验室(桂林理工大学),广西桂林541004

出　　处：《中国科技论文》2023年第11期1221-1229,共9页China Sciencepaper

基　　金：国家自然科学基金资助项目(41961063);广西高校中青年教师科研基础能力提升项目(2022KY0250)。

摘　　要：针对当前室内地图构建过程中未充分考虑可移动障碍物的问题,提出了一种室内障碍物要素图构建方法。首先,利用PointNet++-KPConv获取室内点云的语义信息;随后,采用滤波及体素下采样进行处理;然后,通过随机采样一致性(random sample consensus,RANSAC)算法及欧式聚类对门、墙及室内障碍物点云进行提取,并提出了一种障碍物类型判别方法,将室内障碍物划分为可移动障碍物和不可移动障碍物;最后,将要素点云投影到二维平面构建室内障碍物要素图。结果表明,所提方法可以有效提取室内要素,障碍物类型判别方法的总体准确率达到97.96%,生成的障碍物要素图能够正确地表达室内环境语义信息。To solve the problem that movable obstacles are not fully considered in the current indoor map construction process,an new indoor obstacle element map construction method was proposed.Firstly,PointNet++-KPConv was utilized to obtain the semantic information of indoor point clouds.Secondly,filtering and voxel downsampling were adopted for preprocessing.The point clouds of doors,walls,and indoor obstacles were extracted through the random sample consensus(RANSAC)algorithm and Euclidean clustering.An obstacle type discrimination method to classify the indoor obstacles into movable obstacles and immovable obstacles was proposed.Finally,the elemental point cloud was projected to the two-dimensional plane to construct the indoor obstacle elemental map.The results show that the proposed method can effectively extract indoor elements,the overall accuracy of the obstacle type discrimination method reaches 97.96%,and the generated obstacle element map can correctly express the semantic information of indoor environment.

关 键 词：室内地图 障碍物要素图 障碍物提取 点云语义分割 欧式聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]