摘要
基于Kinect体感技术获取的周围环境点云数据量大,其中点云的边界是重要特征,是机器人导航的重要参数。为获得复杂散乱点云的边界特征,提出了一种基于点云库(PCL)的物体分割以及边缘轮廓提取算法。该算法通过建立散乱点云的kd-tree空间拓扑结构,经直通滤波、表面平滑处理对点云数据进行去噪、填补空洞。由于实际环境包含大量的平面,因此采用基于随机采样算法(RANSAC)可寻找种子点确定平面,进而应用平面分割找出平面上的感兴趣区域,并计算k邻域点的法线夹角,若大于阈值则为边界特征点。为验证算法的有效性,基于机器人操作系统(ROS),通过PCL点云库,快速、准确地对场景中的物体进行分割以及边缘轮廓提取。实验结果表明,所提出的算法能够快速、准确、有效地提取散乱点云的边界。
The data of the surrounding environment point cloud obtained by Kinect somatosensory technology is large, and the boundary of the point cloud is an important feature and parameter for robot navigation. A method of an object segmentation and edge contour ex- traction based on Point Cloud Library （PCL） has been proposed to obtain the boundary feature of complex scattered cloud. By establis- hing the kd-tree space topology structure of scattered points cloud, the point cloud data have been denoised and filled with holes by pass -through filtering and surface smoothing. The real environment contains a large number of planes, thus the random sampling algorithm （RANSAC） has been employed to find the seed point for determination of the plane used to find the region of interest on the plane. Whether the point is the boundary point is judged by the maximum value of angle difference which has been calculated by the normal di- rection with the k -nearest points. The experiment of object segmentation and edge contour extraction with the PCL in the ROS （ Robot Operating System） environment have been performed rapidly and precisely. The experimental results show that this proposed method has quickly, accurately and effectively obtained the boundary of scattered point cloud.
出处
《计算机技术与发展》
2017年第7期83-86,共4页
Computer Technology and Development
基金
天津市科技支撑计划项目(13ZCZDGX01200)
天津市产学研合作项目(14ZCZDSF00025)
天津市863成果转化项目(13RCHZGX01116
14RCHZGX00862)
关键词
散乱点云
KD-TREE
边界特征提取
分割
Scattered point cloud
kd-tree
boundary characteristic extraction
segmentation
