引入平面分割与空间统计分析的室内稠密点云地面提取方法  

Ground Extraction Method of Indoor Dense Point Cloud with Plane Segmentation and Spatial Information

作　　者：王莉敏[1] 史鹏程 WANG Limin;SHI Pengcheng(Geomatics Center of Henan Province,Zhengzhou 450003,China;School of Computer Science,Wuhan University,Wuhan 430072,China)

机构地区：[1]河南省基础地理信息中心,郑州450003 [2]武汉大学计算机学院,武汉430072

出　　处：《遥感信息》2022年第1期25-31,共7页Remote Sensing Information

摘　　要：针对室内场景地面信息提取问题,提出了一种基于平面和空间信息统计分析的室内稠密点云地面提取方法。首先,对输入的稠密点云进行预处理,实现数据降噪及抽稀;其次,利用区域生长分割平面结构,并根据平面位置关系进行合并与分组;然后,采用平面数量、投影面积、天花板与地面、空间距离等多个室内空间信息分析实现地面初提取;最后,采用图像映射、轮廓提取、孔洞填充、点云映射对初提取的地面点云进行优化,得到最终的地面点云。实验表明,该方法可实现不依赖传感器校正坐标系信息,仅使用稠密点云数据作为输入,准确有效地提取地面,准确率达到90%以上,并在一定程度上降低稠密点云的冗余度。Aiming at the problem of extracting ground object information in indoor scene,a ground segmentation method of indoor dense point cloud based on plane and analysis of spatial information is proposed. Firstly,the input dense point cloud is preprocessed to realize denoising and thinning of the data. Secondly,planes are segmented by region growing,and are merged and grouped according to the planar position. And then,spatial information including planar number,projection area,ceiling and ground,and spatial distance are analyzed to realize the initial ground segmentation. Finally,image mapping,contour extraction,hole filling and point cloud mapping are used to optimize the initial segmented ground,and the optimal ground is obtained. Experiments show that ground can be extracted accurately and effectively based on plane and analysis of spatial information just using dense point cloud as input,independent of the corrected coordinate of the sensor. The accuracy reaches over 90%,and the redundancy of the dense point cloud is reduced to some extent.

关 键 词：区域生长 地面提取 平面结构 空间信息 稠密点云 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]