Author Affiliations
1School of Information Engineering, Southwest University of Science and Technology, Mianyang, Sichuan 621010, China2School of Automation and Information Engineering, Sichuan University of Science & Engineering, Zigong, Sichuan 643000, China;show less
Fig. 1. Point cloud rasterization.(a) Original point cloud rasterization; (b) target point cloud rasterization
Fig. 2. Search for optimal R flowchart based on DE
Fig. 3. Comparison of registration results of various algorithms in Feet perspective 1. (a) Unregistered results; (b) CPD algorithm; (c) Scale-ICP algorithm; (d) Go-ICP algorithm; (e) proposed algorithm
Fig. 4. Comparison of registration results of various algorithms in Feet perspective 2. (a) Unregistered results; (b) CPD algorithm; (c) Scale-ICP algorithm; (d) Go-ICP algorithm; (e) proposed algorithm
Fig. 5. Comparison of registration results of various algorithms in Cow perspective 1. (a) Unregistered results; (b) CPD algorithm; (c) Scale-ICP algorithm; (d) Go-ICP algorithm; (e) proposed algorithm
Fig. 6. Comparison of registration results of various algorithms in Cow perspective 2. (a) Unregistered results; (b) CPD algorithm; (c) Scale-ICP algorithm; (d) Go-ICP algorithm; (e) proposed algorithm
Fig. 7. Comparison of registration results of various algorithms for adding noise to Feet. (a) Unregistered results; (b) CPD algorithm; (c) Scale-ICP algorithm; (d) Go-ICP algorithm; (e) proposed algorithm
Fig. 8. Comparison of registration results of various algorithms for adding noise to Cow. (a) Unregistered results; (b) CPD algorithm; (c) Scale-ICP algorithm; (d) Go-ICP algorithm; (e) proposed algorithm
Fig. 9. Comparison of registration results of various algorithms for Feet losing 50% data. (a) Unregistered results; (b) CPD algorithm; (c) Scale-ICP algorithm; (d) Go-ICP algorithm; (e) proposed algorithm
Fig. 10. Comparison of registration results of various algorithms for Feet losing 75% data. (a) Unregistered results; (b) CPD algorithm; (c) Scale-ICP algorithm; (d) Go-ICP algorithm; (e) proposed algorithm
Fig. 11. Comparison of registration results of various algorithms for Cow losing 50% data. (a) Unregistered results; (b) CPD algorithm; (c) Scale-ICP algorithm; (d) Go-ICP algorithm; (e) proposed algorithm
Fig. 12. Comparison of registration results of various algorithms for Cow losing 75% data. (a) Unregistered results; (b) CPD algorithm; (c) Scale-ICP algorithm; (d) Go-ICP algorithm; (e) proposed algorithm
Model | MSE |
---|
View | CPD | Scale-ICP | Go-ICP | Proposed |
---|
Feet | 1 | 2.67×10-5 | 6.79×10-4 | 1.62×10-9 | 4.90×10-9 | | 2 | 1.00×10-3 | 6.79×10-4 | 8.34×10-4 | 1.45×10-7 | Cow | 1 | 9.21×10-5 | 2.83×10-31 | 8.01×10-10 | 2.87×10-7 | | 2 | 1.30×10-3 | 1.50×10-3 | 1.60×10-3 | 4.82×10-8 |
|
Table 1. Comparison of MSE results of different models from different views
Model | Running time/s |
---|
View | CPD | Scale-ICP | Go-ICP | Proposed |
---|
Feet | 1 | 2.51 | 0.90 | 24.88 | 17.96 | | 2 | 3.37 | 0.89 | 24.75 | 17.77 | Cow | 1 | 3.00 | 0.92 | 24.53 | 26.13 | | 2 | 3.05 | 1.27 | 25.49 | 25.69 |
|
Table 2. Comparison of running time of different models from different views
Model | MSE |
---|
CPD | Scale-ICP | Go-ICP | Proposed |
---|
Feet | 3.86×10-5 | 7.02×10-4 | 2.21×10-5 | 4.84×10-5 | Cow | 1.02×10-4 | 1.50×10-3 | 2.61×10-5 | 6.81×10-5 |
|
Table 3. Comparison of MSE results of interference signal added by different models
Model | MSE |
---|
Missing /% | CPD | Scale-ICP | Go-ICP | Proposed |
---|
Feet | 50 | 3.29×10-5 | 6.60×10-3 | 1.69×10-7 | 1.25×10-5 | | 75 | 3.17×10-5 | 2.40×10-2 | 1.67×10-9 | 7.16×10-7 | Cow | 50 | 2.40×10-4 | 1.89×10-2 | 1.90×10-3 | 1.48×10-4 | | 75 | 9.31×10-4 | 2.84×10-1 | 1.80×10-3 | 2.26×10-4 |
|
Table 4. Comparison of MSE results of different models with different sizes of missing data