测绘学报 ›› 2018, Vol. 47 ›› Issue (2): 269-274.doi: 10.11947/j.AGCS.2018.20170493

• • 上一篇    下一篇

双层优化的激光雷达点云场景分割方法

李明磊1, 刘少创2, 杨欢3, 亓晨3   

  1. 1. 南京航空航天大学电子信息工程学院, 江苏 南京 211106;
    2. 中国科学院遥感与数字地球研究所, 北京 100101;
    3. 武汉大学测绘学院, 湖北 武汉 430079
  • 收稿日期:2017-09-01 修回日期:2017-11-29 出版日期:2018-02-20 发布日期:2018-03-02
  • 作者简介:李明磊(1987-),男,博士,讲师,研究方向为数字摄影测量与遥感。E-mail:minglei_li@126.com
  • 基金资助:
    江苏省自然科学基金(BK20170781)

Bilevel Optimization for Scene Segmentation of LiDAR Point Cloud

LI Minglei1, LIU Shaochuang2, YANG Huan3, QI Chen3   

  1. 1. College of Electronic and Information Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China;
    2. Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100101, China;
    3. School of Geodesy and Geomatics, Wuhan University, Wuhan 430079, China
  • Received:2017-09-01 Revised:2017-11-29 Online:2018-02-20 Published:2018-03-02
  • Supported by:
    The Natural Science Foundation of Jiangsu Province of China (No. BK20170781)

摘要: 对激光雷达扫描的非结构化点云进行分割处理,是进行数据组织、重构和信息提取的重要步骤。本文根据点云表面的局部可微的性质,提出了一种递进形式的双层优化分割算法。首先在黎曼几何框架下计算点的拓扑关系和距离度量特性,以k均值聚类的方法获得过分割体素,作为底层分割结果。然后,将点云的体素模式化为节点,构建最小生成树,提取节点的高级特征信息,利用图优化得到对点云细节自适应的区域分割效果。通过真实数据进行验证,并与现有方法比较,证明所提算法的可行性和先进性。

关键词: 点云分割, 黎曼几何, 超体素, 最小生成树, 特征提取

Abstract: The segmentation of point clouds obtained by light detection and ranging (LiDAR) systems is a critical step for many tasks,such as data organization,reconstruction and information extraction.In this paper,we propose a bilevel progressive optimization algorithm based on the local differentiability.First,we define the topological relation and distance metric of points in the framework of Riemannian geometry,and in the point-based level using k-means method generates over-segmentation results,e.g.super voxels.Then these voxels are formulated as nodes which consist a minimal spanning tree.High level features are extracted from voxel structures,and a graph-based optimization method is designed to yield the final adaptive segmentation results.The implementation experiments on real data demonstrate that our method is efficient and superior to state-of-the-art methods.

Key words: point cloud segmentation, Riemannian geometry, super voxel, minimal spanning tree, feature extraction

中图分类号: