• Chinese Journal of Lasers
  • Vol. 48, Issue 16, 1604002 (2021)
Qiqi Li1、2, Xianghong Hua1、2、*, Bufan Zhao1、2、3, Wuyong Tao1、2, and Cheng Li1、2
Author Affiliations
  • 1School of Surveying and Mapping, Wuhan University, Wuhan, Hubei 430079, China
  • 2Disaster Monitoring and Prevention Research Center of Wuhan University, Wuhan, Hubei 430079, China
  • 3Key Laboratory for Digital Land and Resources of Jiangxi Province, East China University of Technology, Nanchang, Jiangxi 330013, China
  • show less
    DOI: 10.3788/CJL202148.1604002 Cite this Article Set citation alerts
    Qiqi Li, Xianghong Hua, Bufan Zhao, Wuyong Tao, Cheng Li. New Method for Plane Segmentation of Indoor Scene Point Cloud[J]. Chinese Journal of Lasers, 2021, 48(16): 1604002 Copy Citation Text show less
    Processing flow of proposed method
    Fig. 1. Processing flow of proposed method
    Projection length model of point cloud
    Fig. 2. Projection length model of point cloud
    Three sets of data used in experiment. (a) data 1; (b) data 2; (c) data 3
    Fig. 3. Three sets of data used in experiment. (a) data 1; (b) data 2; (c) data 3
    Number of stratified point clouds
    Fig. 4. Number of stratified point clouds
    Normal Gaussian map and point cloud before and after normal constraint. (a) Gaussian ball; (b) partial zoom; (c) constrained normal vector; (d) point cloud at maximum layer 1; (e) point cloud at maximum layer 2; (f) constrained point cloud
    Fig. 5. Normal Gaussian map and point cloud before and after normal constraint. (a) Gaussian ball; (b) partial zoom; (c) constrained normal vector; (d) point cloud at maximum layer 1; (e) point cloud at maximum layer 2; (f) constrained point cloud
    Preliminary segmentation results of data 1. (a) Partitioned plane point clouds; (b) floor and ceiling plans; (c) residual point clouds
    Fig. 6. Preliminary segmentation results of data 1. (a) Partitioned plane point clouds; (b) floor and ceiling plans; (c) residual point clouds
    Plane segmentation results of data 1 after model optimization
    Fig. 7. Plane segmentation results of data 1 after model optimization
    Plane segmentation results of point clouds in data 2 and data 3. data 2 plane point cloud, (a) segmentation results, (b) residual point cloud; data 3 plane point cloud, (c) segmentation results, (d) residual point cloud
    Fig. 8. Plane segmentation results of point clouds in data 2 and data 3. data 2 plane point cloud, (a) segmentation results, (b) residual point cloud; data 3 plane point cloud, (c) segmentation results, (d) residual point cloud
    PlaneABD
    Plane 1213.27450.03051971.6564
    Plane 2-2.5988×10-51.7823×10-5-0.0034
    Plane 326.3625-26.9836469.1850
    Plane 4-2.7181×10-61.0147×10-53.3648
    Plane 5-82.5261-0.3624-1122.2972
    Plane 6-498.51250.0325-5334.4432
    Plane 7-281.2597-281.1161-2618.1471
    Plane 80.00161.1324×10-41.1738
    Plane 90.0273-148.163977.4453
    Plane 10-0.2417-93.3212197.1178
    Plane 1125.7364-25.7179454.0566
    Table 1. Parameters of planes in point cloud of data 1 scene
    ParameterPlane 1Plane 2Plane 3Plane 4Plane 5Plane 6Plane 7Plane 8Plane 9Plane 10
    ρ9.2440.00312.4333.36513.59810.7016.5841.1740.5232.112
    Peak interval[9.225,9.275][0.025,0.075][12.425,12.475][3.325,3.375][13.625,13.675][10.675,10.725][6.575,6.625][1.125,1.175][0.475,0.525][2.125,2.175]
    Table 2. Comparison between peak interval of projection length ρ of plane model and projection length of point cloud
    ObjectModeldistance /mAverage distance /mMeasured distance /mDistanceerror /mmRelative accuracy /%
    Ceiling-desktop Desktop-ceiling2.25982.25922.25952.25722.399.89
    Floor-ceilingCeiling-floor3.02253.02213.02233.01804.399.86
    Front wall-back wallBack wall-front wall5.97305.97445.97375.97491.299.97
    Left wall-right wallRight wall-left wall7.75777.75447.75617.75322.999.96
    Table 3. Segmentation precision of plane point cloud
    ObjectCeilingFloorDesktopFront wallBack wallLeft wallRight wall
    Deflection angle/(°)0.0770.0560.1470.0280.1700.1350.103
    Table 4. Deflection angle of planes
    DataNumber of pointsProposed method /sMLESAC method/sImproved 3D-HT method /s
    11145824.76015.68710.291
    240928616.72460.38531.516
    3167243467.997208.213140.930
    Table 5. Segmentation time-consuming of plane point clouds
    Qiqi Li, Xianghong Hua, Bufan Zhao, Wuyong Tao, Cheng Li. New Method for Plane Segmentation of Indoor Scene Point Cloud[J]. Chinese Journal of Lasers, 2021, 48(16): 1604002
    Download Citation