摘要
针对传统多尺度模型对模型点云比较方法(Multiscale Model to Model Cloud Comparison,M3C2)计算法向量与形变量时易受离群点影响的缺点,提出一种基于离群点探测准则的改进算法。首先,在估计关键点法向量时,依据改进离群点探测准则迭代剔除离群点,提高法向量估计的准确性,然后,通过离群点探测剔除圆柱内离群点,最后,结合正态分布加权计算形变量。实验结果表明,相较于M3C2原始算法,改进算法将法向量均方差精度指标提升50%以上,在形变量较大区域可将形变量估值均方差精度指标提高200%以上。改进算法具有更好的适用性和可靠性。
An improved algorithm based on the outlier detection criterion is proposed to address the shortcomings of the traditional Multiscale Model to Model Cloud Comparison(M3C2)method in cal-culating normal vectors and shape variables.First,the normal vector of key points is estimated based on the improved outlier detection criterion to improve the accuracy of the normal vector estima-tion,and then the outliers in the cylinder are removed by outlier detection.Finally,the shape varia-bles are calculated by combining the normal distribution weighting.The experimental results show that compared with the original M3C2 algorithm,the improved algorithm can improve the accuracy of the normal vector by over 50%,and can improve the accuracy of the shape variable evaluation by o-ver 200%in the large shape variable region.The improved algorithm has better applicability and reliability.
作者
王骈臻
夏元平
刘华
舒研鑫
周欣
WANG Pianzhen;XIA Yuanping;LIU Hua;SHU Yanxin;ZHOU Xin(School of Surveying and Mapping Engineering,East China University of Technology,330013,Nanchang,PRC;Wuhan Huace Satellite Technology Co.,Ltd,430000,Wuhan,PRC)
出处
《江西科学》
2023年第5期970-976,984,共8页
Jiangxi Science
基金
国家自然科学基金项目(42174055,49162018)。
关键词
多尺度模型对模型点云比较
离群点探测
主成分分析
点云形变监测
RANSAC
multiscale model to model cloud comparison
outlier detection
principal component a-nalysis
point cloud deformation monitoring
RANSAC
作者简介
王骈臻(1999-),男,硕士研究生,研究方向:点云数据处理;通信作者:夏元平(1982-),男,教授,博士,研究方向:InSAR技术及应用、三维激光扫描技术。E-mail:1478211215@qq.com。
