室内三维点云空间自动划分与规则化方法  

Automatic division and regularization of indoor 3D point cloud space

作　　者：王敬淳 汤圣君 王伟玺 李晓明 李游 谢林甫 朱家松[1] WANG Jingchun;TANG Shengjun;WANG Weixi;LI Xiaoming;LI You;XIE Linfu;ZHU Jiasong(School of Civil and transportation,Shenzhen University,Shenzhen 518061,China;Smart City Research Institute,School of Architecture and Urban Planning,Shenzhen University,Shenzhen 518061,China;Key Laboratory of Urban Natural Resources Monitoring and Simulation of Ministry of Natural Resources,Shenzhen 518061,China;Shenzhen Key Laboratory of Spatial Information Intelligent Sense and Service,Shenzhen 518061,China;Guangdong Key Laboratory of Urban Spatial Information Engineering,Shenzhen 518061,China)

机构地区：[1]深圳大学土木与交通学院,广东深圳518061 [2]深圳大学建筑与城市规划学院智慧城市研究院,广东深圳518061 [3]自然资源部城市自然资源监测与仿真重点实验室,广东深圳518061 [4]深圳市空间信息智能感知与服务重点实验室,广东深圳518061 [5]广东省城市空间信息工程重点实验室,广东深圳518061

出　　处：《测绘通报》2022年第8期93-97,138,共6页Bulletin of Surveying and Mapping

基　　金：国家自然科学基金(41801392;41901329;41971354;41971341);广东省科技计划(2018B020207005);自然资源部城市自然资源监测与仿真重点实验室项目(KF-2019-04-010)。

摘　　要：精准空间划分是实现室内语义建模与拓扑结构重建的重要基础。三维点云作为常用的室内空间数据载体,如何基于三维点云进行室内空间语义信息提取与规则化具有重要意义。本文提出了一种基于形态学分割方法实现室内场景的分割,并结合矢量规则化方法完成分割场景的规则化。首先,基于区域增长算法与线性拟合方法提取空间分割要素,通过平面投影生成二进制影像,进而利用距离变换和分水岭算法完成空间分割;然后,对空间分割要素进行线性拟合,进行室内空间格网划分,采用矢栅叠加方法实现空间要素规则化;最后,通过4组实际场景(包含3组ISPRS数据集及1组实际场景采集数据)进行数据验证。试验结果显示,本文提出的室内空间分割与规则化方法可以准确快速地完成室内空间要素的提取。Accurate spatial division is an important basis for realizing indoor semantic modeling and topology reconstruction.As a commonly used indoor space data carrier, 3 D point cloud is of great significance to extract and normalize semantic information in indoor space based on 3 D point cloud.This paper proposes an indoor scene segmentation based on the morphological segmentation method, and combines the vector regularization method.Firstly, spatial segmentation elements are extracted based on regional growth algorithm and linear fitting method, generate binary image through plane projection, and then use distance transformation and watershed algorithm to complete spatial segmentation. Secondly, linear fit of indoor spatial mesh.This paper verifies 4 sets of actual scene data, including 3 sets of ISPRS data sets and 1 set of actual scene collection data. The test results show that the indoor space segmentation and regularization method proposed in this paper can accurately and quickly complete the extraction of indoor space elements.

关 键 词：三维点云 空间剖分 分水岭 规则化 

分 类 号：P237[天文地球—摄影测量与遥感]