摘要
提出了一种结合全景影像的车载街景点云数据增强方法,首先结合基于密度的聚类方法 DBSCAN(density-based spatial clustering of applications with noise)分割算法和地物典型特征实现点云数据的分类及单体目标提取;然后对单体目标点云,通过构不规则三角网(triangulated irregular network,TIN),逐一进行缺失区域检测及相应边缘提取;最后提出了基于全景影像局部仿射变换的区域增长密集匹配方法,用于生成缺失空洞区域的真实三维点,实现点云数据的增强。实验表明,该方法能够实现车载街景点云数据缺失区域的填补,且点云增强的结果真实、可靠。
In this paper, a method of vehicular point cloud data enhancement with panoramic image is proposed. Firstly, the classification of point cloud data and the extraction of single object interested are realized by combining DBSCAN(density-based spatial clusterig of applications with noise) segmentation algorithm and typical features of objects;then, the missing area is detected and the corresponding edge is extracted one by one for a point cloud of single object;finally, a region growth dense matching based on panoramic image with local affine transform is proposed. The method is used to generate the 3 D point data in the missing area, and to enhance the point cloud data. Experiment results show that this method can fill the missing area of cloud data in streetscape, and the result of point cloud enhancement is real and reliable.
作者
刘亚文
张颖
陈泉
LIU Yawen;ZHANG Ying;CHEN Quan(School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430079,China; The Department of Hitech Industry,Wuhan University,Wuhan 430079,China)
出处
《武汉大学学报(信息科学版)》
EI
CSCD
北大核心
2020年第7期1015-1020,共6页
Geomatics and Information Science of Wuhan University
基金
中央高校基本科研业务费专项资金(2042014Kf0294)。
关键词
车载点云数据
点云聚簇
全景影像匹配
点云缺失
点云数据增强
vehicle point cloud data
point cloud data clustering
panoramic image matching
point cloud data missing
point cloud data enhancement