测绘通报 ›› 2017, Vol. 0 ›› Issue (6): 45-48,93.doi: 10.13474/j.cnki.11-2246.2017.0187

• 学术研究 • 上一篇    下一篇

面向室内场景点云的对象重建

杨泽鑫1, 程效军1,2, 丁琼3, 程小龙1   

  1. 1. 同济大学测绘与地理信息学院, 上海 200092;
    2. 现代工程测量国家测绘地理信息局重点实验室, 上海 200092;
    3. 广东工业大学测绘工程系, 广东 广州 510006
  • 收稿日期:2016-11-15 修回日期:2017-01-23 出版日期:2017-06-25 发布日期:2017-07-03
  • 作者简介:杨泽鑫(1994-),男,硕士生,主要研究方向为点云处理及三维建模。E-mail:zxyang15@163.com
  • 基金资助:
    国家自然科学基金(41671449)

Objects Reconstruction Oriented to Indoor Scene Point Cloud

YANG Zexin1, CHENG Xiaojun1,2, DING Qiong3, CHENG Xiaolong1   

  1. 1. College of Surveying and Geo-Informatics, Tongji University, Shanghai 200092, China;
    2. Key Laboratory of Advanced Engineering Surveying of NASMG, Shanghai 200092, China;
    3. Department of Surveying and Mapping, Guangdong University of Technology, Guangzhou 510006, China
  • Received:2016-11-15 Revised:2017-01-23 Online:2017-06-25 Published:2017-07-03

摘要: 针对当前逆向工程中对象提取及模型重建效率较低的问题,提出了一种面向室内场景点云的对象重建方法。首先构建直通滤波器,采用改进的RANSAC算法剔除非对象点云,然后根据欧氏聚类提取算法分割出各个对象点云,最后基于α-shape理论批量重建出对象模型。试验结果表明,该方法能够从散乱的室内场景点云中快速、自动地重建出代表真实对象的三维模型,具有较高的实用价值。

关键词: 室内场景点云处理技术, 点云分割, 三维重建

Abstract: This paper presents an algorithm of reconstructing objects from indoor scene point cloud for solving the low efficiency problem of object extraction and reconstruction in reverse engineering. The steps of this approach are as follows:firstly, the non-object clouds are removed by using pass-through filter and the improved RANSAC algorithm; then the object clouds are segmented according to Euclidean Cluster Extraction algorithm; finally, the object models are reconstructed based on the α-shape theory. The experimental result shows that the proposed method, which is of high practical value, can quickly and automatically reconstruct the 3D models representing real objects from unorganized indoor scene point clouds.

Key words: indoor scene point cloud processing technology, point cloud segmentation, 3D reconstruction

中图分类号: