摘要
针对点云配准时重叠区域小,难以提取特征,配准精度较低的问题,提出了一种结合改进的随机采样一致性(RANSAC)算法和改进的迭代最近点(ICP)算法的点云配准算法。首先,采用快速点特征直方图(FPFH)描述子对点云进行特征描述;其次通过融合几何一致性,采用改进随机采样一致性的算法,及时删除匹配过程中的误匹配点对,保持着对应点之间的优质关系,使其在低重叠率的点云以及含噪声的点云下也能找到具有对应关系的点,进行点云粗配准;最后针对点云数据量大时ICP配准耗时长的问题,采用KD-Tree搜索,将无序的点云进行有序化排列,进行点云精配准。采用激光雷达扫描的真实点云数据进行实验验证,并与主流点云配准算法进行比较分析。实验结果表明,对于较低重叠率、含噪声的点云,能够快速、精确地求得最优变换,具有较好的配准效果。
Aiming at the problems of low overlap rate,difficult to extract features and low registration accuracy in point cloud registration,a point cloud registration algorithm combining improved random sample consensus(RANSAC)algorithm and improved iterative nearest point(ICP)algorithm is proposed.Firstly,fast point feature histogram(FPFH)descriptor is used to describe the feature of point cloud.Secondly,by fusing geometric consistency and adopting the algorithm of improving random sampling consistency,the mismatched point pairs in the matching process are deleted in time to maintain the high-quality relationship between the corresponding points.The points with corresponding relationship can also be found under the point cloud with low overlap rate and the point cloud with noise for rough registration of point cloud.Finally,aiming at the problem that ICP registration takes a long time when the amount of point cloud data is large,KD-Tree search is used to orderly arrange the disordered point clouds for fine registration.The real point cloud data scanned by lidar are used for experimental verification,and compared with the mainstream point cloud registration algorithm.The experimental results show that for the point cloud with low overlap rate and noise,the optimal transformation can be obtained quickly and accurately,and has a good registration effect.
作者
黄丽婷
林靖宇
卢泉
HUANG Liting;LIN Jingyu;LU Quan(School of Electrical Engineering,Guangxi University,Nanning 530004;School of Automation,Guangdong University of Technology,Guangzhou 510006)
出处
《计算机与数字工程》
2024年第9期2543-2548,2554,共7页
Computer & Digital Engineering
基金
国家自然科学基金项目(编号:61561005)资助。
