Laser & Optoelectronics Progress, Volume. 58, Issue 6, 610008(2021)
Adaptive Point Cloud Reduction Based on Multi Parameter k-Means Clustering
Figures & Tables(12)
Fig. 1. Initial clustering results of bunny models with different k values. (a) k=500; (b) k=1000; (c) k=1500; (d)k=2000
Fig. 3. Feature point detection results of each model under corresponding parameters. (a) Bunny model feature; (b) hippo model feature; (c) golden bucket model feature
Fig. 4. Reduced model diagram of bunny model under various reduction ratios. (a) Reduction ratio is 0.1; (b) reduction ratio is 0.2; (c) reduction ratio is 0.5
Fig. 5. Reduced model diagram of hippo model under various reduction ratios. (a) Reduction ratio is 0.1; (b) reduction ratio is 0.2; (c) reduction ratio is 0.5
Fig. 6. Reduced model diagram of golden bucket model under various reduction ratios. (a) Reduction ratio is 0.1; (b) reduction ratio is 0.2; (c) reduction ratio is 0.5
Fig. 7. Results of bunny model reduction by different methods. (a) Random reduction; (b) curvature sampling; (c) uniform grid; (d) our method
Fig. 8. Results of hippo model reduction by different methods. (a) Random reduction; (b) curvature sampling; (c) uniform grid;(d) our method
Fig. 9. Results of golden bucket model reduction by different methods. (a) Random reduction; (b) curvature sampling; (c) uniform grid; (d) our method
Table 1. Comparison results of different initialization cluster center selection methods in bunny model
View tableTable 1. Comparison results of different initialization cluster center selection methods in bunny model
Method k value Number of iterations Mean value of objective function DBI Run time /s Max Mean Min Our method 500 52 52 52 0.622311 0.872 3.986 1000 31 31 31 0.318098 0.883 4.333 1500 31 31 31 0.214407 0.908 5.325 2000 20 20 20 0.161110 0.895 5.451 Traditional k-means clustering method 500 74 62 52 0.624288 0.864 5.952 1000 61 43 29 0.319547 0.918 12.449 1500 39 33 26 0.215288 0.939 19.842 2000 34 25 19 0.162071 0.933 30.369
Table 2. Comparison of standard deviation and surface area change rate under different reduction rates
View tableTable 2. Comparison of standard deviation and surface area change rate under different reduction rates
Model Simplification Number of data Standard deviation Change rate of surface area /% Bunny 0.1 3733 0.000465 2.0601 0.2 7140 0.000396 1.3272 0.5 18278 0.000351 0.6114 Hippo 0.1 3722 0.230015 16.7310 0.2 6583 0.155388 15.7344 0.5 11569 0.110577 14.6062 Golden bucket 0.1 6735 4.463796 5.1303 0.2 13662 2.998224 3.5730 0.5 33325 1.935787 3.1996
Table 3. Comparison of standard deviation and surface area change rate under different simplification methods
View tableTable 3. Comparison of standard deviation and surface area change rate under different simplification methods
Method Bunny Hippo Golden bucket Standard deviation Change rate of surface area /% Standard deviation Change rate of surface area /% Standard deviation Change rate of surface area /% Random 0.000415 1.6097 0.265509 16.3719 3.684469 4.8828 Curvature 0.001101 1.9231 2.365784 28.1071 3.590388 12.6053 Grids 0.000379 1.2535 1.158725 16.1662 4.299177 5.8593 Our method 0.000396 1.3272 0.155388 15.7344 2.998224 3.5730
Tools
Get Citation
Copy Citation Text
Wang Jianqiang, Fan Yanguo, Li Guosheng, Yu Dingfeng. Adaptive Point Cloud Reduction Based on Multi Parameter k-Means Clustering[J]. Laser & Optoelectronics Progress, 2021, 58(6): 610008
Paper Information
Category: Image Processing
Received: Jul. 22, 2020
Accepted: --
Published Online: Mar. 11, 2021
The Author Email: Dingfeng Yu (z18010014@s.upc.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20