• Laser & Optoelectronics Progress
  • Vol. 56, Issue 1, 011203 (2019)
Ming Liu1, Qin Shu1、*, Yunxiu Yang2, and Fei Yuan2
Author Affiliations
  • 1 School of Electrical Engineering and Information, Sichuan University, Chengdu, Sichuan 610065, China
  • 2 Southwest Institute of Technical Physics, Chengdu, Sichuan 610041, China
  • show less
    DOI: 10.3788/LOP56.011203 Cite this Article Set citation alerts
    Ming Liu, Qin Shu, Yunxiu Yang, Fei Yuan. Three-Dimensional Point Cloud Registration Based on Independent Component Analysis[J]. Laser & Optoelectronics Progress, 2019, 56(1): 011203 Copy Citation Text show less
    Flow chart of proposed algorithm
    Fig. 1. Flow chart of proposed algorithm
    Point cloud of Bunny. (a) Independent component feature graph of original point cloud; (b) independent component feature graph of unregistered point cloud
    Fig. 2. Point cloud of Bunny. (a) Independent component feature graph of original point cloud; (b) independent component feature graph of unregistered point cloud
    Registration of Bunny.(a) Pre-registration point cloud; (b) registration by the proposed algorithm; (c) registration by scale-ICP algorithm; (d) registration by CPD algorithm; (e) registration by classic ICP algorithm
    Fig. 3. Registration of Bunny.(a) Pre-registration point cloud; (b) registration by the proposed algorithm; (c) registration by scale-ICP algorithm; (d) registration by CPD algorithm; (e) registration by classic ICP algorithm
    Point cloud of Elephant. (a) Independent component feature graph of original point cloud; (b) independent component feature graph of unregistered point cloud
    Fig. 4. Point cloud of Elephant. (a) Independent component feature graph of original point cloud; (b) independent component feature graph of unregistered point cloud
    Registration of Elephant. (a) Pre-registration point cloud; (b) registration by the proposed algorithm; (c) registration by scale-ICP algorithm; (d) registration by CPD algorithm; (e) registration by classic ICP algorithm
    Fig. 5. Registration of Elephant. (a) Pre-registration point cloud; (b) registration by the proposed algorithm; (c) registration by scale-ICP algorithm; (d) registration by CPD algorithm; (e) registration by classic ICP algorithm
    Point cloud registration after adding noise. (a) Pre-registration point cloud; (b) registration by the proposed algorithm; (c) registration by scale-ICP algorithm; (d) registration by CPD algorithm; (e) registration by classic ICP algorithm
    Fig. 6. Point cloud registration after adding noise. (a) Pre-registration point cloud; (b) registration by the proposed algorithm; (c) registration by scale-ICP algorithm; (d) registration by CPD algorithm; (e) registration by classic ICP algorithm
    Registration of spare part. (a) Pre-registration point cloud; (b) registration by the proposed algorithm; (c) registration by scale-ICP algorithm; (d) registration by CPD algorithm; (e) registration by classic ICP algorithm
    Fig. 7. Registration of spare part. (a) Pre-registration point cloud; (b) registration by the proposed algorithm; (c) registration by scale-ICP algorithm; (d) registration by CPD algorithm; (e) registration by classic ICP algorithm
    Point cloud of spare part. (a) Independent component feature graph of original point cloud; (b) independent component feature graph of unregistered point cloud
    Fig. 8. Point cloud of spare part. (a) Independent component feature graph of original point cloud; (b) independent component feature graph of unregistered point cloud
    Point cloudAlgorithmTime /sMSE /mm
    BunnyICP53.4760.195
    CPD algorithm28.4731.098×10-10
    Scale-ICP algorithm6.6958.439×10-15
    Proposed algorithm0.8311.625×10-28
    ElephantICP38.4050.014
    CPD algorithm21.0551.1577×10-8
    Scale-ICP algorithm5.7247.559×10-16
    Proposed algorithm0.4732.246×10-30
    Table 1. Comparison of point cloud registrations with different algorithms
    AlgorithmTime /sMSE /mm
    ICP67.3520.2687
    CPD algorithm37.4270.1021
    Scale-ICP algorithm11.9460.1301
    Proposed algorithm0.9740.1025
    Table 2. Comparison of point cloud registrations after adding noise
    AlgorithmTime /sMSE /mm
    ICP196.3520.385
    CPD algorithm124.4273.0605×10-14
    Scale-ICP algorithm102.8573.4667×10-11
    Proposed algorithm2.5349.019×10-22
    Table 3. Comparison of point cloud registrations of spare part
    Ming Liu, Qin Shu, Yunxiu Yang, Fei Yuan. Three-Dimensional Point Cloud Registration Based on Independent Component Analysis[J]. Laser & Optoelectronics Progress, 2019, 56(1): 011203
    Download Citation