基于ISS-3DSC特征的点云配准方法  

Point cloud registration method based on ISS-3DSC characteristics

作　　者：王一 张鑫淼 王莹 苏皓 WANG Yi;ZHANG Xinmiao;WANG Ying;SU Hao(College of Electrical Engineering,North China University of Technology,Tangshan 063210,China;Tangshan Advanced Testing and Control Technology Key Laboratory,Tangshan 063210,China)

机构地区：[1]华北理工大学电气工程学院,河北唐山063210 [2]唐山市先进测试与控制技术重点实验室,河北唐山063210

出　　处：《中国测试》2025年第2期62-67,80,共7页China Measurement & Test

基　　金：河北省高等学校科学技术研究项目(ZD2022114);唐山市科技计划项目(21130212C)。

摘　　要：点云配准是三维数据处理的重要步骤,但传统关键点的4点一致集算法配准过程中存在计算复杂度较高、配准用时较长等问题。为此,提出一种基于ISS-3DSC特征改进的K4PCS算法(keypoint-based 4-points congruent sets,K4PCS)。将下采样后的点云数据经内部形状描述子算法(intrinsic shape signatures,ISS)提取关键点后,由三维形状上下文特征(3D shape context,3DSC)描述形成ISS-3DSC特征点,并将ISS-3DSC特征点作为K4PCS算法粗配准的输入数据,再采用基于中值距离改进的迭代最近点算法(iterative closest point,ICP)完成精配准。实验结果表明,本文算法在公开数据集上总用时缩短15.45%~76.54%,且效果更明确,具有一定实用性。Point cloud registration is an important step in 3D data processing,but the traditional Keypoint-Based 4-Points Congruent Sets(K4PCS)algorithm registration process suffers from higher computational complexity and longer registration time.For this reason,a K4PCS algorithm is proposed based on the improved ISS-3DSC features.The downsampled point cloud data are extracted from the keypoints by the Intrinsic Shape Signatures(ISS)algorithm,and then described by the 3D Shape Context(3DSC)to form the ISS-3DSC feature points,and the ISS-3DSC feature points are used as the K4PCS algorithm coarse-registration input data.This paper proposes an ICP algorithm based on median distance improvement to achieve fine registration.The experimental results show that the algorithm reduces the total time by 15.45%-76.54%.on the public dataset,and the effect is more explicit,which is practical to some extent.

关 键 词：点云配准 三维形状上下文特征 关键点的四点一致集 迭代最近点 

分 类 号：TB9[一般工业技术—计量学] TN249[机械工程—测试计量技术及仪器]