随机采样一致算法中不同约束条件的比较研究
A Comparative Study of Different Constraints in Random Sampling Consensus Algorithm
陶武勇 1周梦韦 1花向红 2徐少平 1肖艳阳1
作者信息
- 1. 南昌大学数学与计算机学院,江西南昌,330031
- 2. 武汉大学测绘学院,湖北武汉,430079
- 折叠
摘要
随机采样一致(random sampling consensus,RANSAC)算法被广泛用于点云配准中估计转换参数.近年来,许多学者提出在RANSAC算法中添加约束条件,提高配准精度和缩短时间,但对于这些约束条件的约束效果和它们之间的关联关系尚不清楚.本文对4种约束条件进行比较研究,将4种约束条件组成16种组合,分别采用两种描述符为点云建立初始配对集,然后使用不同约束条件下的RANSAC算法计算转换参数,比较配准精度和计算效率.实验对比分析这些组合,分析不同组合的效果,给出不同约束条件的优缺点.
Abstract
The RANSAC algorithm has been widely used in point cloud registration to estimate transformation parame-ters. In recent years,many researchers propose to improve registration accuracy and shorten time by adding constraints to the RANSAC algorithm. However,the effects of the con-straints and their relationships are not yet clear. This paper performs a comparison study for four constraints and combines them into sixteen combinations. In the experiments,two de-scriptors are respectively used to establish the initial correspon-dences set for point clouds,and then the RANSAC algorithm with different constraints is used to calculate the transforma-tion parameters. The registration accuracy and computational efficiency are compared. By experiments,these These combi-nations are also compared and analyzed in the experiment. The effects of different combinations are compared,and the advantages and disadvantages of different constraints are giv-en.
关键词
点云配准/随机采样一致/约束条件/局部形状描述符Key words
point cloud registration/RANSAC/constraints/local shape descriptor引用本文复制引用
基金项目
国家自然科学基金(42271452)
国家自然科学基金(62162043)
国家自然科学基金(62102174)
出版年
2024
NETL
NSTL
万方数据