面向服务机器人三维地图创建的大规模点云分割  

3D Unorganized Large-scale Point Clouds Segmentation for Service Robot

作　　者：邓成呈 曹其新[1] 李彰植[1] 

机构地区：[1]上海交通大学机械系统与振动国家重点实验室,上海200240

出　　处：《机电一体化》2012年第6期21-24,共4页Mechatronics

基　　金：教育部博士点基金资助(项目号:20090073110037);863重点项目(2006AA040203)

摘　　要：为了使服务机器人能在室内环境中识别物体并且利用点云自动创建三维VRML地图,针对室内场景中大规模且无规则的点云分割,提出了一种结合RANSAC和聚类的分割算法。该算法充分考虑到室内场景的特征,利用RANSAC提取大平面点云,并在点云当中去除地面等平面环境中的点云数据,最后使用聚类分割物体。此算法充分利用了室内环境的大平面特征和空间上分离特征,有效避免了无规则点云的分割和室内复杂环境导致的分割和运算困难的问题。不同室内环境中的对比实验比较清晰地分割出环境中的主要物体,得到了比较满意的效果。To enable service robot to localize itself in the environment, it＇s quite essential to perceive and reconstruct the environment. As to the unorganized large-scale point clouds segmentation, this paper introduced a novel algorithm which integrates the RANSAC algorithm and clustering methods together to solve the 3D point cloud segmentation. In this process, after filtering the raw point cloud, points that belong the ground and roof, as well as walls, are extracted by RANSAC algorithm, and then objects are segmented after data clustering. The algorithm proposed in this paper solves the problem of unorganized large-scale point cloud segmentation. Finally, the algorithm are verified in different indoor environment.

关 键 词：服务机器人 地图创建 点云分割 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]