K-近邻长方体的点云特征提取压缩算法  

A point cloud compression algorithm based on K-neighborhood cuboid

作　　者：宋敏峰 贾东振[1] 郭俊文 何秀凤[1] SONG Minfeng;JIA Dongzhen;GUO Junwen;HE Xiufeng(School of Earth Sciences and Engineering,Hohai University,Nanjing 211100,China;North Information Control Research Academy Group Co.,Ltd.,NORINCO Group,Nanjing 211100,China)

机构地区：[1]河海大学地球科学与工程学院,南京211100 [2]中国兵器工业集团北方信息控制研究院集团有限公司,南京211100

出　　处：《测绘科学》2019年第10期93-100,共8页Science of Surveying and Mapping

基　　金：国家自然科学基金重点项目(41830110);国家自然科学基金项目(41474001)

摘　　要：针对海量点云数据存在大量冗余问题,该文提出基于K-近邻长方体的点云压缩算法。利用目标点的K近邻在非特征点云与特征点云之间的不同分布特性,基于该文算法将点云集合分为特征及非特征点集。该方法先对目标点近邻点进行坐标转换并构建K-近邻长方体,建立压缩准则,对长方体进行扁平程度筛选,结合分段采样去除大量冗余点及少量密集特征点,实现保留原始特征的点云压缩。该文方法涉及K、α、采样率βall3个参数,在实验分析中,采用体积偏差、表面积偏差和Hausdorff距离对该文方法涉及的3个参数进行精度影响分析,结果表明,该方法能保留大量原始特征,在最优K值条件下βall为0.4,α为0.9,此时体积偏差百分比为0.27%,表面积偏差百分比为0.5%,具有较高的压缩精度。Aiming at data redundancy in point cloud,this paper proposed a simplification algorithm based on K-neighborhood cuboid.Point cloud data was divided into feature point set and non feature point set by the different characteristics distributed between the feature and non feature point cloud part based on the K-neighborhood cuboid algorithm.The K-neighborhood cuboid which built after transforming the coordinate of the nearest neighbor point were filtered according to the property by the compression criterion,and then the two types of point set were compressed in different sampling rate,in order that compressed point could retain enough geometrical characteristics.Three parameters:K,α,and sample rate βall were involved in our algorithm,and which were analyzed in detail by volume deviation surface area deviation and Hausdorff distance in the experiment.Under the optimal Kvalue,the βall was 0.4 and the α was 0.9,the volume deviation percentage was 0.27%,and the surface area deviation percentage wa s0.5%.It showed that our compression algorithm could preserve the details of the reconstruction model and get higher compression precision.

关 键 词：K-近邻长方体 点云 数据压缩 特征提取 K近邻 

分 类 号：P234[天文地球—摄影测量与遥感]