基于ISS-SHOT+3D-NDT的点云配准算法研究  

Research on Point Cloud Registration Algorithm Based on ISS-SHOT+3D-NDT

作　　者：马国鹏 宁殿民[1] 卞艳[1] MA Guopeng;NING Dianmin;BIAN Yan(School of Civil Engineering,University of Science and Technology Liaoning,Anshan,Liaoning 114051,China)

机构地区：[1]辽宁科技大学土木工程学院,辽宁鞍山市114051

出　　处：《矿业研究与开发》2021年第5期175-181,共7页Mining Research and Development

基　　金：国家重点研发计划项目(2016YFC0801602).

摘　　要：针对以往点云数据在配准处理过程中,存在粗配准匹配率低、配准时间长、精配准精度不高等问题,提出了ISS-SHOT与3D-NDT相结合的点云配准算法。首先,运用内部形态描述子算法(ISS)对下采样后的点云数据提取特征点,对提取的特征点用方向直方图描述子(SHOT)进行描述,并利用相似度函数将对应的特征点进行匹配;再采用Lmeds算法删除匹配错误的特征点,并完成2片点云数据的粗配准,获得较好的初始配准位置与姿态;最后,采用3D-NDT算法将粗配准后的点云数据进行空间体素化处理,运用概率分布函数完成点云数据的精确配准。结果表明,与其他点云数据配准算法相比,ISS-SHOT+3D-NDT算法的均方根误差最小、配准精度最高,特征点匹配率较高且运行时间短。In the previous point cloud data registration process,there are some problems,such as low coarse registration rate,long registration time and poor precision.Aiming at these problems,a point cloud registration algorithm based on ISS-SHOT and 3 D-NDT was proposed.Firstly,the feature points was extracted by the intrinsic shape signatures(ISS)descriptor algorithm from the point cloud data after down-sampling,and then were described by signature of histograms of orientations(SHOT)descriptor.Then,the corresponding feature points were matched by similarity function.Secondly,Lmeds algorithm is used to eliminate the mismatched feature points,and the coarse registration for two pieces of point cloud data was completed to obtain better position and attitude of initial registration.Finally,3 D-NDT algorithm was adopted to conduct spatial voxelization for the point cloud data after coarse registration,and probability distribution function was used to complete the accurate registration for point cloud data.The results show that,compared with other point cloud data registration algorithms,ISS-SHOT+3 D-NDT algorithm has the least root mean square error,the highest registration accuracy,a higher feature point matching rate and shorter running time.

关 键 词：ISS算法 方向直方图描述子 3D-NDT 点云配准 

分 类 号：TD679[矿业工程—矿山机电]