• Laser & Optoelectronics Progress
  • Vol. 58, Issue 24, 2415004 (2021)
Xuchun Zhang1, Hongjun Zhou2, Jinjin Zheng1、*, and Yi Jin1、**
Author Affiliations
  • 1Department of Precision Machinery and Precision Instrumentation, University of Science and Technology of China, Hefei, Anhui 230026, China
  • 2National Synchrotron Radiation Laboratory, University of Science and Technology of China, Hefei, Anhui 230027, China
  • show less
    DOI: 10.3788/LOP202158.2415004 Cite this Article Set citation alerts
    Xuchun Zhang, Hongjun Zhou, Jinjin Zheng, Yi Jin. Point Cloud Registration Based on Multi-Scale Feature and Point Distance Constraint[J]. Laser & Optoelectronics Progress, 2021, 58(24): 2415004 Copy Citation Text show less
    Principle of the SPFH
    Fig. 1. Principle of the SPFH
    Neighbor scale of center point p
    Fig. 2. Neighbor scale of center point p
    Matching principle of key points
    Fig. 3. Matching principle of key points
    Point cloud registration result of our algorithm. (a) Original point cloud; (b) key points; (c) matching points and corresponding relations; (d) results of registration
    Fig. 4. Point cloud registration result of our algorithm. (a) Original point cloud; (b) key points; (c) matching points and corresponding relations; (d) results of registration
    Fragment of the Bunny point cloud model. (a) bun000 and bun045; (b) bun090 and bun045
    Fig. 5. Fragment of the Bunny point cloud model. (a) bun000 and bun045; (b) bun090 and bun045
    Registration result of the Bunny point cloud. (a) bun000 and bun045; (b) bun090 and bun045
    Fig. 6. Registration result of the Bunny point cloud. (a) bun000 and bun045; (b) bun090 and bun045
    Experimental results of office point cloud registration. (a) Original point cloud; (b) 4PCS; (c) Super-4PCS; (d) our algorithm
    Fig. 7. Experimental results of office point cloud registration. (a) Original point cloud; (b) 4PCS; (c) Super-4PCS; (d) our algorithm
    Serial numberPoint cloud modelSegments of modelNumber of points
    1Bunnybun00040256
    bun04540097
    2ArmadilloarmadilloSide_6023404
    armadilloSide_4520647
    3BuddhahappySideRight_33668890
    happySideRight_31257499
    4DragondragonStandRight_33643467
    dragonStandRight_28824573
    Table 1. Size of partial segments of point cloud models in The Stanford 3D Scanning Repository dataset
    Point cloud modelSegments of modelNumber of points
    Bunnybun00040256
    bun04540097
    bun09030379
    Table 2. Size of partial point cloud segments of Bunny model
    Serial numberAlgorithmRoot mean square distance error
    bun000 and bun045bun045 and bun090
    1ICP0.0009000.017943
    24PCS0.0048020.010270
    3Super-4PCS0.0038030.010086
    4GICP0.0012290.010159
    5Ours-CR0.0009840.007196
    6Ours-FR0.0008960.006465
    Table 3. Registration results of different algorithms unit: mm
    Serial numberAlgorithmRoot mean square distance error /mm
    14PCS0.139283
    2Super-4PCS0.135356
    3ours0.094396
    Table 4. Experimental results of registration algorithm for the point cloud of office environment
    Xuchun Zhang, Hongjun Zhou, Jinjin Zheng, Yi Jin. Point Cloud Registration Based on Multi-Scale Feature and Point Distance Constraint[J]. Laser & Optoelectronics Progress, 2021, 58(24): 2415004
    Download Citation