《计算机应用研究》|Application Research of Computers

基于假设检验匹配约束的点云配准算法研究

Research of point cloud registration algorithm based on hypothesis test matching constraints

免费全文下载 (已被下载 次)  
获取PDF全文
作者 江旭,耿楠,张志毅,胡少军
机构 西北农林科技大学 a.信息工程学院;b.农业农村部农业物联网重点实验室;c.陕西省农业信息感知与智能服务重点实验室,陕西 杨凌 712100
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2021)01-062-0305-06
DOI 10.19734/j.issn.1001-3695.2019.07.0569
摘要 针对点云配准中效率低、误差大、抗噪性弱等问题,提出了一种改进的基于t检验的迭代最近点(T-ICP)算法。在初始配准阶段,采用统计分析对源点云和目标点云中的离群点进行标记并提取非离群点,然后采用主成分分析法(PCA)计算非离群源点云和非离群目标点云之间的变换矩阵,并将变换矩阵应用于源点云。在精配准阶段,以迭代最近点(ICP)算法作为基本框架,通过对候选点对的邻域距离分布进行t检验来剔除错误点对,并采用均匀分布策略来搜索点对,保证点云的完整形态配准。实验结果表明,相较于迭代最近点算法以及近两年一些改进的配准算法,该算法在效率和精度上分别提高了10%~50%和4%~40%,并具有较好的鲁棒性。
关键词 点云配准; 主成分分析; 迭代最近点; 邻域距离分布; t检验; 均匀分布
基金项目 陕西省重点研发计划资助项目(2019ZDLNY07-06-01)
本文URL http://www.arocmag.com/article/01-2021-01-062.html
英文标题 Research of point cloud registration algorithm based on hypothesis test matching constraints
作者英文名 Jiang Xu, Geng Nan, Zhang Zhiyi, Hu Shaojun
机构英文名 a.College of Information Engineering,b.Key Laboratory of Agricultural Internet of Things for Ministry of Agriculture & Rural Affairs,c.Shaanxi Key Laboratory of Agricultural Perception & Intelligent Service,Northwest A&F University,Yangling Shaanxi 712100,China
英文摘要 Aiming at the problems of low efficiency, large error and weak anti-noise ability in point cloud registration, this paper proposed an improved iterative closest point registration algorithm based on t test(test-iterative closest point, T-ICP). In initial registration, this paper used statistical analysis to mark outliers in point cloud and extracted non-outliers. Then, it used principal component analysis(PCA) to calculate the transformation matrix between the non-outlier source point cloud and the non-outlier target point cloud, and the transformation matrix could transform source point cloud to target point cloud. In fine registration, this paper used iterative closest point(ICP) algorithm as the basic framework and introduced t test and uniform distribution. T test could analyze the neighborhood distance distribution of candidate point pairs and eliminate wrong point pairs. Uniform distribution as the strategy of searching point pairs could ensure complete morphological registration of point cloud. Experimental results show that the proposed algorithm improves the efficiency and accuracy by 10%~50% and 4%~40%, respectively, and has better robustness, compared with ICP and some improved registration algorithms in the last two years.
英文关键词 point cloud registration; PCA; ICP; neighborhood distance distribution; t test; uniform distribution
参考文献 查看稿件参考文献
 
收稿日期 2019/7/2
修回日期 2019/9/1
页码 305-310
中图分类号 TP391.41
文献标志码 A