基于多级体素的支柱及支持装置点云自动提取方法

以三维激光扫描点云数据为研究对象,首先利用八叉树构建多级体素;其次,通过分析支柱及支持装置的空间、几何特征生成种子体素;然后,利用区域生长方法,制定生长规则,分离出单个支柱及支持装置;最后,以单个支柱及支持装置为处理单元,采用渐进式的提取策略细化提取点云,获取单个支柱及支持装置的有效、精确数据,实现支柱及支持装置点云的自动提取,并进行试验验证。结果表明:该方法不仅能完整、准确地提取所有支柱及支持装置,且单个支柱及支持装置点云数目提取的平均完整度和正确率分别达到94.08%和94.48%,验证该方法的有效性和优异性;通过构建多级体素,不仅提高了邻域搜索的速度,还提高了单个支柱及支持装置的点云提取精度。

Method for Automatic Extraction of Pillar and Supporting Device from Point Cloud Data Based on Multi-Level Voxel

The point cloud data of 3D laser-scanning is researched in the paper.Firstly,multi-level voxels are constructed by the octree method.Secondly,the seed voxels are generated by analyzing the spatial and geometric features of the pillars and supporting devices.Thirdly,the growth rules are formulated by using the regional growth method,and the single pillar and supporting device are separated.Finally,the effective and accurate data extracted by a single pillar and supporting device are obtained by using the progressive extraction strategy to refine and extract the point cloud,in which the single pillar and supporting device are treated as one processing unit.Thus the automatic extraction of the point cloud of the pillar and supporting device is realized,and the experimental verification is carried out.Results show that this method not only can extract all the pillars and supporting devices completely and accurately,but also extract the number of single pillar and supporting device with an average integrity and accuracy of 94.08% and 94.48% respectively,which proves the effectiveness and superiority of the method.Furthermore,by constructing multi level voxels,not only the speed of neighborhood search is improved,but also the point cloud extraction accuracy of single pillar and supporting device is improved.