室内SLAM点云的快速分割研究

Automated Segmentation for Indoor SLAM Point Cloud Based on Region Growing Algorithm

建筑物内部结构和环境复杂,由此产生的噪声等影响使其无法直接运用于室外建模中已经成熟的点云分割算法。为降低噪声带来的影响,本文提出利用直方图统计法,分别对点云进行Z轴方向和X-Y轴方向上的直方图统计,从而分割出地板面,天花板面以及墙面的"候选点"。以K-D树构建空间数据索引,计算点云中各点的法向量以及曲率,将"候选点"中面与面相交处曲率"突变"的点去除,利用区域生长算法分割出建筑物的地板面、墙面以及天花板面。以NavVis公司的M3三维激光扫描车获取的室内SLAM点云数据,对本文方法检验,实验结果表明该方法能有效地降低噪声带来的影响,并且可以对平面点云数据进行分割提取。

Considering the noise coming with the complicated inner structure and environment of the buildings,the existing outdoor point cloud segmentation algorithms cannot satisfy 3D modeling well. In this paper,we applied histogram statistics into 3D modeling to reduce the influence of the noise. In our method,applying histogram statistics to the directions of Z axis and X-Y axis respectively to get the candidate points of the ground,roof and wall. Calculating the normal vectors and curvature of point cloud via the spatial data index by K-D tree algorithm. And in our algorithm,we implement the region growing algorithm to get area of the ground,roof and wall after removing the candidate points of curvature mutation in the intersection of surfaces. The indoor SLAM point cloud data collected by the M3 3D laser scanning vehicle has been adopted to evaluate the proposed method. The experimental results show that the proposed method can reduce the influence of the noise significantly,and our method can segment and extract the point cloud well.