• Infrared and Laser Engineering
  • Vol. 50, Issue 12, 20210115 (2021)
Wentao Cui, Weidong Jiao, and Yanli Pang
Author Affiliations
  • Key Laboratory of Intelligent Signal and Image Processing, Civil Aviation University of China, Tianjin 300300, China
  • show less
    DOI: 10.3788/IRLA20210115 Cite this Article
    Wentao Cui, Weidong Jiao, Yanli Pang. SVS-NLMS point cloud registration algorithm based on geometric algebra[J]. Infrared and Laser Engineering, 2021, 50(12): 20210115 Copy Citation Text show less
    Flow chart of proposed algorithm
    Fig. 1. Flow chart of proposed algorithm
    Experiment results of cube data set simulation. (a) Raw data; (b) Result of SAC-IA+ICP algorithm registration; (c) Result of GA-SVSNLMS algorithm registration
    Fig. 2. Experiment results of cube data set simulation. (a) Raw data; (b) Result of SAC-IA+ICP algorithm registration; (c) Result of GA-SVSNLMS algorithm registration
    Convergence curves of each algorithm in geometric algebraic space of cube dataset. (a) Convergence curves of error function; (b) Convergence curves of cost function
    Fig. 3. Convergence curves of each algorithm in geometric algebraic space of cube dataset. (a) Convergence curves of error function; (b) Convergence curves of cost function
    Experiment results of bunny data set simulation. (a) Raw data; (b) Result of SAC-IA+ICP algorithm registration; (c) Result of GA-SVSNLMS algorithm registration
    Fig. 4. Experiment results of bunny data set simulation. (a) Raw data; (b) Result of SAC-IA+ICP algorithm registration; (c) Result of GA-SVSNLMS algorithm registration
    Convergence curves of each algorithm in geometric algebraic space of bunny dataset. (a) Error function curve and cost function curve of GA-SVSNLMS; (b) Cost function convergence curve of each algorithm in geometric algebraic space
    Fig. 5. Convergence curves of each algorithm in geometric algebraic space of bunny dataset. (a) Error function curve and cost function curve of GA-SVSNLMS; (b) Cost function convergence curve of each algorithm in geometric algebraic space
    Simulation experiment results of each data set under Gaussian noise. (a) Raw data of cube; (b) Result of SAC-IA+ICP algorithm registration of cube; (c) Result of GA-SVSNLMS algorithm registration of cube; (d) Raw data of bunny; (e) Result of SAC-IA+ICP algorithm registration of bunny; (f) Result of GA-SVSNLMS algorithm registration of bunny
    Fig. 6. Simulation experiment results of each data set under Gaussian noise. (a) Raw data of cube; (b) Result of SAC-IA+ICP algorithm registration of cube; (c) Result of GA-SVSNLMS algorithm registration of cube; (d) Raw data of bunny; (e) Result of SAC-IA+ICP algorithm registration of bunny; (f) Result of GA-SVSNLMS algorithm registration of bunny
    AlgorithmRMSE/mmConvergence speed/times
    ICP${\rm{2}}.{\rm{5079}} \times {10^{ - 2}}$1000
    SAC-IA+ICP${\rm{2}}.2{\rm{483}} \times {10^{ - 2}}$1000
    GA-LMS$1.2617 \times {10^{ - 8}}$750
    GA-NLMS( $\;\beta = 0$) $1.2684 \times {10^{ - 8}}$140
    GA-NLMS( $\;\beta = {\rm{1}}$) $1.2679 \times {10^{ - 8}}$175
    GA-SVSNLMS$1.8852 \times {10^{ - 8}}$70
    Table 1. Running accuracy and convergence speed of different algorithms
    AlgorithmRMSE/mmConvergence speed/times
    ICP${\rm{5}}.{\rm{4046}} \times {10^{ - {\rm{3}}}}$200
    SAC-IA+ICP${\rm{4}}.{\rm{4687}} \times {10^{ - {\rm{3}}}}$200
    GA-LMS${\rm{2}}{\rm{.9595}} \times {10^{ - {\rm{3}}}}$210
    GA-NLMS( $\;\beta = 0$) ${\rm{2}}{\rm{.9443}} \times {10^{ - {\rm{3}}}}$55
    GA-NLMS( $\;\beta = {\rm{1}}$) ${\rm{2}}{\rm{.7620}} \times {10^{ - {\rm{3}}}}$85
    GA-SVSNLMS${\rm{2}}{\rm{.6658}} \times {10^{ - {\rm{3}}}}$40
    Table 2. Running accuracy and convergence speed of different algorithms
    Algorithmcubebunny
    RMSE/mmConvergence speed/timesRMSE/mmConvergence speed/times
    ICP${\rm{2}}.{\rm{284\;7}} \times {10^{ - {\rm{1}}}}$1000${\rm{1}}{\rm{.286\;4}} \times {10^{ - 2}}$200
    SAC-IA+ICP${\rm{3}}{\rm{.328\;6}} \times {10^{ - 2}}$1000${\rm{6}}{\rm{.398\;4}} \times {10^{ - {\rm{3}}}}$200
    GA-LMS${\rm{5}}{\rm{.227\;6}} \times {10^{ - {\rm{3}}}}$750${\rm{6}}{\rm{.015\;9}} \times {10^{ - {\rm{3}}}}$210
    GA-NLMS( $\;\beta = 0$) ${\rm{5}}{\rm{.603\;7}} \times {10^{ - {\rm{3}}}}$140${\rm{6}}{\rm{.019\;4}} \times {10^{ - {\rm{3}}}}$55
    GA-NLMS( $\;\beta = {\rm{1}}$) ${\rm{5}}{\rm{.576\;8}} \times {10^{ - {\rm{3}}}}$175${\rm{6}}{\rm{.005\;6}} \times {10^{ - {\rm{3}}}}$85
    GA-SVSNLMS${\rm{5}}{\rm{.237\;7}} \times {10^{ - {\rm{3}}}}$70${\rm{5}}{\rm{.841\;5}} \times {10^{ - {\rm{3}}}}$40
    Table 3. Running accuracy and convergence speed of different algorithms under Gaussian noise
    Wentao Cui, Weidong Jiao, Yanli Pang. SVS-NLMS point cloud registration algorithm based on geometric algebra[J]. Infrared and Laser Engineering, 2021, 50(12): 20210115
    Download Citation