通过对比学习优化深度拟合的点云法向估计
Optimization of Point Cloud Normal Estimation for Deep Fitting via Contrastive Learning
DOI: 10.12677/PM.2023.137210, PDF,    科研立项经费支持
作者: 史路冰*, 聂明辉, 张 杰：辽宁师范大学数学学院，辽宁 大连
关键词: 深度学习；对比学习；法向估计；Deep Learning； Contrastive Learning； Normal Estimation
摘要: 近年来，以当前点的邻域作为输入，利用深度网络估计逐点权重进行加权最小二乘曲面拟合的算法，在法向估计上取得了当前领先的结果。本文算法在此基础之上，引入对比学习的思想，即先通过真实法向的差异构造三元组，再根据构造的三元组创建三元组损失，用以对邻域内的每点权重进行约束，从而提高权重的估计质量，获得更优的法向估计结果。实验结果表明，本文算法能够对已有的算法有较好的提高，在PCPNet数据集上的均方误差平均值为11.71。
Abstract: In recent years, algorithms that use the neighborhood of the current point as input and use deep networks to estimate point-wise weights for weighted least squares surface fitting have achieved leading results in normal estimation. On this basis, the algorithm in this paper introduces the idea of contrastive learning, which first constructs triplets based on the differences in real normal di-rections, and then creates triplet losses based on the constructed triplets to constrain the weights of each point in the neighborhood, thereby improving the estimation quality of weights and obtaining better normal estimation results. The experimental results show that the proposed algorithm can significantly improve existing algorithms, with an average mean square error of 11.71 on the PCPNet dataset.
文章引用：史路冰, 聂明辉, 张杰. 通过对比学习优化深度拟合的点云法向估计[J]. 理论数学, 2023, 13(7): 2037-2043. https://doi.org/10.12677/PM.2023.137210

参考文献

[1] Hoppe, H., DeRose, T., Duchamp, T., et al. (1992) Surface Reconstruction from Unorganized Points. ACM SIGGRAPH Computer Graphics, 26, 71-78. [Google Scholar] [CrossRef] 
[2] Pauly, M., Keiser, R., Kobbelt, L.P., et al. (2003) Shape Modeling with Point-Sampled Geometry. ACM Transactions on Graphics, 22, 641-650. [Google Scholar] [CrossRef] 
[3] Amenta, N. and Bern, M. (1999) Surface Reconstruction by Voronoi Filtering. Discrete & Computational Geometry, 22, 481-504. [Google Scholar] [CrossRef] 
[4] Dey, T.K. and Goswami, S. (2006) Provable Surface Reconstruction from Noisy Samples. Computational Geometry Theory, 35, 124-141. [Google Scholar] [CrossRef] 
[5] Feng, H.Y. and Ouyang, D. (2005) On the Normal Vector Estimation for Point Cloud Data from Smooth Surfaces. Computer-Aided Design, 37, 1071-1079. [Google Scholar] [CrossRef] 
[6] Jones, T.R., Durand, F. and Zwicker, M. (2004) Normal Im-provement for Point Rendering. IEEE Computer Graphics and Applications, 24, 53-56. [Google Scholar] [CrossRef] 
[7] Boulch, A. and Marlet, R. (2016) Deep Learning for Robust Normal Estimation in Unstructured Point Clouds. Computer Graphics Forum, 35, 281-290. [Google Scholar] [CrossRef] 
[8] Lu, D.N., Lu, X.Q., Sun, Y.X., et al. (2020) Deep Feature-Preserving Normal Estimation for Point Cloud Filtering. Computer-Aided Design, 125, Article ID: 102860. [Google Scholar] [CrossRef] 
[9] Guerrero, P., Kleiman, Y., Ovsjanikov, M., et al. (2018) PCPNet: Learning Local Shape Properties from Raw Point Clouds. Computer Graphics Forum, 37, 75-85. [Google Scholar] [CrossRef] 
[10] Zhou, J., Huang, H., Liu, B., et al. (2020) Normal Estimation for 3D Point Clouds via Local Plane Constraint and Multi-Scale Selection. Computer-Aided Design, 129, Article ID: 102916. [Google Scholar] [CrossRef] 
[11] Qi, C.R., Su, H., Mo, K., et al. (2017) PointNet: Deep Learning on Point Sets for 3d Classification and Segmentation. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, 21-26 July 2017, 652-660. [Google Scholar] [CrossRef] 
[12] Zhu, R.S., Liu, Y., Dong, Z., et al. (2021) AdaFit: Rethinking Learning-Based Normal Estimation on Point Clouds. Proceedings of the IEEE/CVF International Conference on Computer Vision, Montreal, 10-17 October 2021, 6118-6127. [Google Scholar] [CrossRef] 
[13] Ben-Shabat, Y. and Gould, S. (2020) DeepFit: 3d Surface Fitting via Neural Network Weighted Least Squares. European Conference on Computer Vision, Glasgow, 23-28 August 2020, 20-34. [Google Scholar] [CrossRef] 
[14] Zhang, J., Cao, J.J., Zhu, H.R., et al. (2021) Geometry Guided Deep Surface Normal Estimation. Computer-Aided Design, 142, Article ID: 103119. [Google Scholar] [CrossRef] 

为你推荐