首页 > 论文 > 光学学报 > 40卷 > 18期(pp:1815001--1)

采用空间投影的深度图像点云分割

Depth Image Point Cloud Segmentation Using Spatial Projection

  • 摘要
  • 论文信息
  • 参考文献
  • 被引情况
  • PDF全文
分享:

摘要

点云分割是点云处理的一个关键环节,其分割质量决定了目标测量、位姿估计等任务的精确与否。提出了一种采用空间投影的深度图像(RGB-D)点云分割方法,在分析了相机模型、RGB-D数据特征以及图像阈值与目标点云关系的基础上,建立靶标坐标系与点云区域的模型,进一步地结合靶标坐标系和图像阈值,把点云变换至靶标坐标系以突出目标区域、弱化背景区域,并用图像形态学处理所投影的像素值以及分割图像以获得所对应的点云区域。建立3种测试场景以获得3组不同的点云数据,采用4种方法对点云进行分割对比,其中采用空间投影的方法能获得较高的点云分割质量;对空间投影中的膨胀元素、数值与分割质量的关系进行测试分析,结果表明了采用空间投影的方法对RGB-D点云分割的有效性和可行性。

Abstract

Point cloud segmentation is a key step in point cloud processing, and its segmentation quality determines the accuracy of target measurement, pose estimation, and other tasks. This paper proposes a method of depth image (RGB-D) point cloud segmentation using spatial projection. Based on the camera model, RGB-D data characteristics, and the relationship between the image threshold and the target point cloud, a target coordinate system and point cloud regions are established. Further, based on the target coordinate system and the image threshold, the point cloud is transformed to the target coordinate system to highlight the target region and weaken the background region. Also, the projected pixel values are processed by image morphology and the corresponding point cloud region is obtained by segmenting the image. Finally, three test scenarios are established to acquire three different groups of point cloud data, and four methods are adopted to segment and compare point clouds. The spatial projection based method can obtain better point cloud segmentation quality. The relationship among the expansion element, numerical value, and segmentation quality is tested and analyzed. The results show that the spatial projection method is effective and feasible for RGB-D point cloud segmentation.

广告组1 - 空间光调制器+DMD
补充资料

中图分类号:TP391

DOI:10.3788/AOS202040.1815001

所属栏目:机器视觉

基金项目:广东省重点领域研发计划;

收稿日期:2020-04-28

修改稿日期:2020-06-11

网络出版日期:2020-09-01

作者单位    点击查看

郭清达:华南理工大学电子与信息学院, 广东 广州 510641
全燕鸣:华南理工大学机械与汽车工程学院, 广东 广州 510641

联系人作者:全燕鸣(meymquan@scut.edu.cn)

备注:广东省重点领域研发计划;

【1】Nguyen A, Le B. 3D point cloud segmentation: a survey . [C]//2013 6th IEEE Conference on Robotics, Automation and Mechatronics (RAM), November 12-15, 2013, Manila, Philippines. New York: IEEE. 2013, 225-230.

【2】Ou X L, Kuang X L, Ni W Y. Summarization on 3D scattered point cloud segmentation [J]. Journal of Hunan University of Technology. 2010, 24(5): 45-49.
欧新良, 匡小兰, 倪问尹. 三维散乱点云分割技术综述 [J]. 湖南工业大学学报. 2010, 24(5): 45-49.

【3】Besl P J, Jain R C. Segmentation through variable-order surface fitting [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence. 1988, 10(2): 167-192.

【4】Koh J, Suk M, Bhandarkar S M. A multi-layer Kohonen''''s self-organizing feature map for range image segmentation . [C]//IEEE International Conference on Neural Networks, March 28-April 1,1993, San Francisco, CA, USA, New York: IEEE. 1993, 4725511.

【5】Fischler M A, Bolles R C. Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography [J]. Communications of the ACM. 1981, 24(6): 381-395.

【6】Lu R R, Zhu F, Wu Q X, et al. A fast segmenting method for scenes with stacked plate-shaped objects [J]. Acta Optica Sinica. 2019, 39(4): 0412003.
鲁荣荣, 朱枫, 吴清潇, 等. 一种板型物体混叠场景的快速分割算法 [J]. 光学学报. 2019, 39(4): 0412003.

【7】Fan X H, Xu G L, Li W L, et al. Target segmentation method for three-dimensional LiDAR point cloud based on depth image [J]. Chinese Journal of Lasers. 2019, 46(7): 0710002.
范小辉, 许国良, 李万林, 等. 基于深度图的三维激光雷达点云目标分割方法 [J]. 中国激光. 2019, 46(7): 0710002.

【8】Krizhevsky A, Sutskever I, Hinton G. ImageNet classification with deep convolutional neural networks[C]//Proceedings of the 26th Annual Conference on Neural Information Processing Systems 2012, December 3-6, 2012, South Lake Tahoe, NV, USA. Cambridge: , 2012, 1097-1105.

【9】Klokov R, Lempitsky V. Escape from cells: deep kd-networks for the recognition of 3D point cloud models . [C]//2017 IEEE International Conference on Computer Vision (ICCV), October 22-29, 2017, Venice, Italy. New York: IEEE. 2017, 863-872.

【10】Wang P S, Liu Y, Guo Y X, et al. O-CNN: octree-based convolutional neural networks for 3D shape analysis [J]. ACM Transactions on Graphics. 2017, 36(4): 72.

【11】Riegler G, Ulusoy A O, Geiger A. OctNet: learning deep 3D representations at high resolutions . [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), July 21-26, 2017, Honolulu, HI, USA. New York: IEEE. 2017, 6620-6629.

【12】Charles R Q, Su H, Mo K C, et al. PointNet: deep learning on point sets for 3D classification and segmentation . [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), July 21-26, 2017, Honolulu, HI, USA. New York: IEEE. 2017, 77-85.

【13】Qi C R, Yi L, Su H, et al. -06-07)[2020-04-28] . 2017, org/abs/1706: 02413.Qi C R, Yi L, Su H, et al. -06-07)[2020-04-28] . 2017, org/abs/1706: 02413.

【14】Wu Y L, Quan Y M, Guo Q D. Research on calibration of structured light system based on local homography matrix [J]. Science Technology and Engineering. 2016, 16(13): 186-189, 195.
武彦林, 全燕鸣, 郭清达. 基于局部单应性矩阵的结构光系统标定研究 [J]. 科学技术与工程. 2016, 16(13): 186-189, 195.

【15】Otsu N. A threshold selection method from gray-level histograms [J]. IEEE Transactions on Systems, Man, and Cybernetics. 1979, 9(1): 62-66.

引用该论文

Guo Qingda,Quan Yanming. Depth Image Point Cloud Segmentation Using Spatial Projection[J]. Acta Optica Sinica, 2020, 40(18): 1815001

郭清达,全燕鸣. 采用空间投影的深度图像点云分割[J]. 光学学报, 2020, 40(18): 1815001

您的浏览器不支持PDF插件,请使用最新的(Chrome/Fire Fox等)浏览器.或者您还可以点击此处下载该论文PDF