Infrared and Laser Engineering, Volume. 53, Issue 6, 20240120(2024)
3D surface reconstruction method through neural-pull based on edge enhancement
Figures & Tables(18)
Fig. 1. Neural-pull in 2D surface: Pulling surrounding points to implicit surface, where,
Fig. 4. In the input point cloud (green point in the figure), 8
Fig. 7. Qualitative comparison with other surface reconstruction methods
Fig. 8. Qualitative comparison of sparse point clouds reconstruction
Fig. 10. Taking element2 as an example, when
Table 1. Edge enhancement input effectiveness analysis (
\begin{document}$ \mathrm{CD}\times10^2 $\end{document} View tableView in ArticleTable 1. Edge enhancement input effectiveness analysis (
\begin{document}$ \mathrm{CD}\times10^2 $\end{document} Method 1 2 3 4 5 6 7 NP 0.78 1.05 0.75 0.70 0.58 0.75 0.78 NPEE 0.77 0.81 0.66 0.69 0.55 0.53 0.76
Table 2. CD comparison with other surface reconstruction methods
View tableView in ArticleTable 2. CD comparison with other surface reconstruction methods
Method Metrics Element1 Element2 Element3 DeepSDF acc com CD 0.102 0.114 0.216 0.050 0.151 0.201 0.057 0.108 0.165 Onsurface acc com CD 0.014 0.019 0.033 0.006 0.007 0.013 0.006 0.008 0.014 NP acc com CD 0.003 0.008 0.011 0.003 0.005 0.008 0.004 0.007 0.011 NPEE (ours) acc com CD 0.005 0.007 0.012 0.002 0.005 0.007 0.003 0.005 0.008
Table 3. CD comparison with other surface reconstruction methods in sparse point clouds
View tableView in ArticleTable 3. CD comparison with other surface reconstruction methods in sparse point clouds
Method Metrics alligator bimba buddha bunny dragon DeepSDF acc com CD 0.037 0.086 0.123 0.072 0.082 0.154 0.037 0.061 0.098 0.066 0.087 0.153 0.052 0.114 0.196 Onsurface acc com CD 0.010 0.017 0.027 0.008 0.007 0.015 0.008 0.006 0.014 0.012 0.011 0.023 0.013 0.011 0.024 NP acc com CD 0.034 0.013 0.047 0.007 0.010 0.017 0.013 0.011 0.024 0.027 0.022 0.049 0.028 0.016 0.044 NPEE (ours) acc com CD 0.013 0.011 0.024 0.005 0.005 0.010 0.007 0.006 0.013 0.011 0.010 0.021 0.015 0.013 0.028
Table 4. CD comparison of reconstruction results with different values of
\begin{document}$ \sigma $\end{document} View tableView in ArticleTable 4. CD comparison of reconstruction results with different values of
\begin{document}$ \sigma $\end{document} $ \sigma \times {10^2} $ Metrics 1 3 5 7 element2 acc×102 com×102 CD×102 0.28 0.51 0.79 0.24 0.48 0.73 0.29 0.51 0.81 1.50 1.54 3.05 element3 acc×102 com×102 CD×102 0.42 0.59 1.06 0.42 0.53 0.95 0.44 0.61 1.06 0.47 0.63 1.11
Table 5. CD comparison of reconstruction results by residual structure
View tableView in ArticleTable 5. CD comparison of reconstruction results by residual structure
Model name Metrics Without residual structure NPEE alligator $ \mathrm{CD}\times10^2 $ 1.463 0.812 bimba $ \mathrm{CD}\times10^2 $ 0.668 0.548 buddha $ \mathrm{CD}\times10^2 $ 0.772 0.761
Table 6. CD comparison with other surface reconstruction methods in scanned model reconstruction
View tableView in ArticleTable 6. CD comparison with other surface reconstruction methods in scanned model reconstruction
Method Metrics anchor children gargoyle DeepSDF acc com CD 0.102 0.114 0.216 0.045 0.079 0.124 0.070 0.094 0.164 Onsurface acc com CD 0.007 0.009 0.016 0.005 0.005 0.011 0.009 0.007 0.016 NP acc com CD 0.005 0.008 0.013 0.006 0.005 0.011 0.007 0.005 0.012 NPEE (ours) acc com CD 0.006 0.006 0.012 0.005 0.005 0.010 0.006 0.005 0.011
Tools
Get Citation
Copy Citation Text
Baochang XU, Yihao WANG, Weiwei HAO, Shixuan YIN, Wei WANG, Yafei LI. 3D surface reconstruction method through neural-pull based on edge enhancement[J]. Infrared and Laser Engineering, 2024, 53(6): 20240120
Paper Information
Category:
Received: Mar. 13, 2024
Accepted: --
Published Online: Jul. 31, 2024
The Author Email: WANG Yihao (ricepudding12138@163.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20