摘要
在对现有的点云数据分割和拟合算法进行深入研究的基础上,指出现有算法的不足.充分利用扫描线数据自身固有的特点,提出了新的算法.对RANSAC算法进行了改进,改进后的算法既具有较好的抗差性能,又在计算效率上较现有的抗差算法有了较大的提高,且能够得到更准确的提取结果和更合理的扫描点分隔归属.提出了平面拟合计算过程中拟合直线段端点的定权算法,解决了现有算法中由于拟合直线段端点权重不同无法直接参与平面拟合计算的问题.提出了完整的细碎平面剔除规则.实例证明,利用该算法能够取得较好的点云数据拟合平面自动提取结果.
A profound research reveals the defects of the existing point cloud segmentation and fitting algorithms. The paper presents new algorithms according to the intrinsic characteristics of scanned line data. Random sample consensus （RANSAC） algorithm is modified, and a more robust and efficient algorithm is obtained with the better extraction result and more rational segmentation result. A weight determining algorithm for ends of fitted line segments used in plane fitting is proposed, which solves the problem in traditional algorithms that ends of line segments can＇t be used in plane fitting directly because of the different weights. Besides, a set of integral invalid plane removing algorithm is proposed. A case study show that better plane extraction results of point cloud are achieved with the proposed algorithm.
出处
《同济大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2009年第9期1250-1255,共6页
Journal of Tongji University:Natural Science
