参数曲面理想人造点云生成算法

doi:10.3778/j.issn.1002-8331.2308-0415

摘要/Abstract

摘要： 人造点云广泛用于科学研究、机器学习、动画制作和教学等领域。介绍一种参数曲面人造3D点云生成算法，该算法能根据曲面参数方程及指定的参数范围和噪音强度生成点云，使之沿曲面方向服从同等密度分布，曲面法向服从同一标准差的蓝噪音模式化分布（如正态分布），从而模拟理想3D数字化设备的拍摄结果。算法设计基于参数速度调节、曲率补偿、蓝噪音采样等技术，参数速度调节使得点云不因参数速度变化造成密度差异，曲率补偿技术使得生成的点云在曲面尖锐部位也能获得比较均匀的法向噪音效果。实验结果显示生成的点云呈现密度、厚度、噪音均无明显差别的“均匀”视觉效果和量化指标，并且各向同性。该算法作为3D数字化设备的低成本虚拟替代物，可应用于曲面重建测试和机器学习等领域。

关键词: 人造点云, 参数曲面, 参数曲线, 等分布

Abstract: Synthetic point clouds are widely used in science research, machine learning, animation making, and pedagogue. In this article a 3D point cloud generation algorithm for parametric surfaces is introduced. The algorithm can generate qualified point clouds distributed along a surface with specified parameter equations, paramenter ranges, and noise strengths. The generated results have a uniform volume density along the surface, blue noises in normal directions with an identical standard deviation, thus they artificially simulate scanned data obtained with ideal 3D acquisition equipments. The algorithm is based on a parameter velocity adjustment method, a curvature compensation technique and a blue noise sampling technique, which can eliminate the density differences arised from paramenter velocity changes and make the noises at high curvature parts be pattern-wise distributed with an identical standard deviation. The experimental results show that the density, thickness and noise magnitude of the generated point clouds are uniform in visual effects as well as in quantitative metrics, and that the point clouds are isotropic. As the algorithm is perceived as a low-cost virtual 3D equipment, it will be applied in surface reconstruction testing or machine learning.

Key words: synthetic point cloud, parametric surface, parametric curve, iso-distribution

胡明晓, 何才透. 参数曲面理想人造点云生成算法[J]. 计算机工程与应用, 2023, 59(20): 219-227.

HU Mingxiao, HE Caitou. Qualified Synthetic Point Cloud Generation Algorithm for Parametric Surfaces[J]. Computer Engineering and Applications, 2023, 59(20): 219-227.

参考文献

[1] DONG Z，CHEN J，LIN H.Topology-controllable implicit surface reconstruction based on persistent homology[J].Computer-Aided Design，2022，150：103308.
[2] XIE F，ZHAO J W，JU F.The point cloud collection of the incisor teeth of beaver and re-construction of its curved surface[J].Advanced Materials Research，2012，1597（426）：387-390.
[3] UGGLA G，HOREMUZ M.Towards synthesized training data for semantic segmentation of mobile laser scanning point clouds：generating level crossings from real and synthetic point cloud samples[J].Automation in Construction，2021，130（4）：1-12.
[4] CHEN Y，WANG Z，ZOU L，et al.Quasi-balanced self-training on noise-aware synthesis of object point clouds for closing domain gap[C]//Proceedings of 2022 European Conference on Computer Vision（ECCV2022），2022：1-13.
[5] WONG K M，WONG T T.Blue noise sampling using an N-body simulation-based method[J].Visual Computer，2017，33（6/7/8）：823-832.
[6] JIANG M，ZHOU Y，WANG R，et al.Blue noise sampling using an SPH-based method[J].ACM Transactions on Graphics，2015，34（6）：1-11.
[7] LU Q，XIAO M，LU Y，et al.Attention-based dense point cloud reconstruction from a single image[J].IEEE Access，2019，7：137420-137431.
[8] LIN Y，WANG Y，LI Y，et al.Single view point cloud generation via unified 3D prototype[C]//Proceedings of 35th AAAI Conference on Artificial Intelligence/33rd Conference on Innovative Applications of Artificial Intelligence/11th Symposium on Educational Advances in Artificial Intelligence，2021：2064-2072.
[9] WEN T，KWASNIEWSKI T.Phase noise simulation and modeling of ADPLL by SystemVerilog[C]//2008 IEEE International Behavioral Modeling and Simulation Workshop，2008：29-34.
[10] ZHOU L，DU Y，WU J.3D shape generation and completion through point-voxel diffusion[C]//2021 IEEE/CVF International Conference on Computer Vision（ICCV），2021：5806-5815.
[11] LI Y，LIU H，TAO Y，et al.Reasoning mechanism：an effective data reduction algorithm for online point cloud selective sampling of sculptured surfaces[J].Computer-Aided Design，2019，113：48-61.
[12] LIN Y J，BENZIGER R R，HABIB A.Planar-based adaptive down-sampling of point clouds[J].Photogrammetric Engineering & Remote Sensing，2016，82（12）：955-966.
[13] HANKE T，SCHAERMANN A，GEIGER M，et al.Generation and validation of virtual point cloud data for automated driving systems[C]//2017 IEEE 20th International Conference on Intelligent Transportation Systems（ITSC），2017：1-6.
[14] WANG F，ZHUANG Y，GU H，et al.Automatic generation of synthetic LiDAR point clouds for 3D data analysis[J].IEEE Transactions on Instrumentation and Measurement，2019，68（7）：2671-2673.
[15] MONTLAHUC J，SHAH G A，POLETTE A，et al.As-scanned point clouds generation for virtual reverse engineering of CAD assembly models[J].Computer-Aided Design and Applications，2019，16（6）：1171-1182.
[16] CHAUDHURY A，BOUDON F，GODIN C.3D plant phenotyping：all you need is labelled point cloud data[C]//Computer Vision-ECCV 2020 Workshops，Glasgow，UK，2020.Berlin，Heidelberg：Springer，2020：244-260.
[17] CORSINI M，CIGNONI P，SCOPIGNO R.Efficient and flexible sampling with blue noise properties of triangular meshes[J].IEEE Transactions on Visualization and Computer Graphics，2012，18（6）：914-924.
[18] KANG C L，LU T N，ZONG M M，et al.Point cloud smooth sampling and surface reconstruction based on moving least squares[J].ISPRS-International Archives of the Photogrammetry，Remote Sensing and Spatial Information Sciences，2020：145-151.
[19] HUANG T，CHEN J，ZHANG J，et al.Fast point cloud sampling network[J].Pattern Recognition Letters，2022，164：216-223.
[20] PATRIKALAKIS N M，MAEKAWA T.Shape interrogation for computer aided design and manufacturing[M].Berlin，Heidelberg：Springer-Verlag，2002.
[21] YU H，ZHANG J，JIAO Z.Geodesics on point clouds[J].Mathematical Problems in Engineering，2014：860136.
[22] DUDA R O，HART P E，STOCK D G.Pattern classification[M].2nd ed.New York：John Wiley & Sons Inc，2001：114-124.