摘要
针对三维点云中心骨架提取问题,提出一种基于改进的自适应k均值聚类预分割引导的点云骨架提取算法.首先,将输入点云体素化,利用八叉树算法覆盖输入点云并下采样实现点云化简;其次,在采样点中自适应选取初始聚类中心对点云进行区域划分,并颜色标记;最后,在区域分割的引导下应用L_(1)-中值骨架提取算法实现点云骨架的提取.该算法主要针对L_(1)-中值算法可重复性差、易丢失细节等缺点进行了改进,并且对输入点云的质量以及形状的几何或拓扑信息,都没有严格的先验要求,可以直接应用到未经任何预处理、含有噪声或离群点的初始扫描点云上.展示了从多种不规则点云提取的骨架结果,包括矮小植物、人体动作等.与传统算法相比,该算法具有高准确率、强鲁棒性、强学习扩展能力等优点.
Aiming at the extraction problem of three-dimensional point cloud center skeleton,a point cloud skeleton extraction algorithm based on improved adaptive k-means clustering pre-segmentation guidance is proposed.Firstly,the input point cloud was voxelized and the octree algorithm was used to cover the input point cloud and to simplify the point cloud.Secondly,the initial clustering center is selected adaptively from the sampling points to divide the region of the point cloud,and the color is marked.At last,under the guidance of region segmentation,L_(1)-medial skeleton extraction algorithm is applied to extract the point cloud skeleton.This algorithm is mainly aimed at the poor repeatability of L_(1)-median algorithm,easy to lose details and other shortcomings,and there are no strict prior requirements for the quality and geometric or topological information of the input point cloud,which can be directly applied to the initial scanning point cloud without any preprocessing,containing noise or outliers.This paper presents the results of skeleton extracted from a variety of irregularly shaped point clouds,including dwarf plants and human movements.Compared with the traditional algorithm,this algorithm has the advantages of high accuracy,strong robustness and strong learning and expansion ability.
作者
鲁斌
范晓明
LU Bin;FAN Xiao-Ming(Department of Computer,North China Electric Power University,Baoding 071000)
出处
《自动化学报》
EI
CAS
CSCD
北大核心
2022年第8期1994-2006,共13页
Acta Automatica Sinica
