改进邻域漂移的多假设检验点云降噪

doi:10.19678/j.issn.1000-3428.0067894

摘要/Abstract

摘要： 在获得三维点云数据时,由于仪器、环境、算法等原因,不可避免地会使获得的点云数据带有噪声,而点云数据所携带的噪声将影响后续的点云处理效果。为了抑制三维点云数据中的噪声,同时保留其不同程度的特征,提出一种改进邻域漂移的多假设点云降噪方法。首先,利用邻域点和法向双张量投票法进行特征描述;接着,利用非参数估计构建转移概率函数,并使用核回归方法对新的采样点进行权重计算,在此基础上运用粒子滤波实现邻域漂移,经过多次迭代得到非局部邻域;然后,通过多假设检验法确定不同特征点处的多个法向,并通过加权平均获得最终法向;最后,采用多假设检验的方法分别对特征点和非特征点进行滤波,产生多个候选点并使用目标函数进行优化,从而对点云模型进行降噪。利用所提方法与RIMLS、EAR、L1、PointNet方法对相同噪声模型进行恢复,并对恢复模型与原模型进行误差分析,结果表明,所提方法的平均降噪精度相比RIMLS、EAR和L1方法分别提高了38.1%、41.3%和12.4%,与PointNet相比约降低2.9%,但是所提方法无须进行数据库训练且可调参。

关键词: 点云降噪, 张量投票, 邻域漂移, 非参数估计, 核回归, 多假设检验

Abstract: In 3D point-cloud data acquisition, the instrument, environment, algorithm, and other factors inevitably introduce noise into the data. The noise carried by point-cloud data affects the subsequent processing of the point cloud. To suppress the noise of 3D point-cloud data while preserving the different characteristics, a multi-hypothesis point cloud noise reduction method with an improved neighborhood drift is proposed. First, the neighborhood points and normal double-tensor voting method are used to describe the features. Subsequently, the transfer probability function is constructed using nonparametric estimation, and the weights of the new sampling points are calculated using the kernel regression method, whereby neighborhood drift is realized using a particle filter, and the nonlocal neighborhood is obtained after several iterations. The multiple normal directions of points with different features are then determined by the multi-hypothesis testing method, and the final normal directions are obtained by weighted average. Finally, feature and non-feature points are filtered using the multi-hypothesis testing method, and the point cloud model is denoised by generating multiple candidate points and using the objective function for optimization. To restore the noise model, the method described in this paper was used alongside Robust Implicit Moving Least Squares (RIMLS), Edge-Aware Resampling (EAR), L1, and PointNet methods. The error between the restored and original data was analyzed. The experimental results show that the average noise reduction accuracy obtained using the proposed method was 38.1%, 41.3%, and 12.4% higher than the accuracies obtained using RIMLS, EAR, and L1, respectively. Compared with PointNet, the average noise reduction accuracy was lower by approximately 2.9% using the proposed method; however, database training and parameter adjustment are not required.

Key words: point cloud noise reduction, tensor voting, neighborhood drift, nonparametric estimation, kernel regression, multi-hypothesis test

中图分类号:

TP391

时志鹏, 冯肖维, 赵一平. 改进邻域漂移的多假设检验点云降噪[J]. 计算机工程, 2024, 50(6): 276-286.

SHI Zhipeng, FENG Xiaowei, ZHAO Yiping. Improved Multi-Hypothesis Test Point Cloud Noise Reduction for Neighborhood Drift[J]. Computer Engineering, 2024, 50(6): 276-286.

https://www.ecice06.com/CN/Y2024/V50/I6/276

参考文献

[1] HAN X F, JIN J S, WANG M J, et al. A review of algorithms for filtering the 3D point cloud[J]. Signal Processing:Image Communication, 2017, 57:103-112.
[2] 冯肖维,姜晨,何敏,等.三维距离图像基于特征估计的自适应平滑[J].光学精密工程, 2019, 27(12):2693-2701. FENG X W, JIANG C, HE M, et al. Adaptive smoothing for three-dimensional range image based on feature estimation[J]. Optics and Precision Engineering, 2019, 27(12):2693-2701.(in Chinese)
[3] 陈赫.点云法向估计与匹配技术的研究[D].大连:大连理工大学, 2018. CHEN H. Research on point cloud normal estimation and alignment[D].Dalian:Dalian University of Technology, 2018.(in Chinese)
[4] LEVIN D. Mesh-independent surface interpolation[M]//BRUNNETT G, HAMANN B, MVLLER H, et al. Geometric modeling for scientific visualization. Berlin,Germany:Springer,2004:37-49.
[5] ALEXA M, BEHR J, COHEN-OR D, et al. Computing and rendering point set surfaces[J]. IEEE Transactions on Visualization and Computer Graphics, 2003, 9(1):3-15.
[6] GUENNEBAUD G, GROSS M. Algebraic point set surfaces[J]. ACM Transactions on Graphics,2007,26(3):1-9.
[7] ÖZTIRELI A C, GUENNEBAUD G, GROSS M. Feature preserving point set surfaces based on non-linear kernel regression[J]. Computer Graphics Forum, 2009, 28(2):493-501.
[8] REN Y J, LI T Z, XU J K, et al. Overall filtering algorithm for multiscale noise removal from point cloud data[J]. IEEE Access, 2021, 9:110723-110734.
[9] SUN Y J, SCHAEFER S, WANG W P. Denoising point sets via L0 minimization[J]. Computer Aided Geometric Design, 2015, 35/36:2-15.
[10] HE L, SCHAEFER S. Mesh denoising via L0 minimization[J]. ACM Transactions on Graphics,2013,32(4):1-8.
[11] LAI Y K, ZHOU Q Y, HU S M, et al. Robust feature classification and editing[J]. IEEE Transactions on Visualization and Computer Graphics, 2007, 13(1):34-45.
[12] LEAL E, SANCHEZ-TORRES G, BRANCH J W. Sparse regularization-based approach for point cloud denoising and sharp features enhancement[J]. Sensors, 2020, 20(11):3206.
[13] LIPMAN Y, COHEN-OR D, LEVIN D. Data-dependent MLS for faithful surface approximation[C]//Proceedings of the 5th Eurographics Symposium on Geometry Processing. New York,USA:ACM Press,2007:59-67.
[14] HUANG H, WU S H, GONG M L, et al. Edge-aware point set resampling[J]. ACM Transactions on Graphics,2013,32(1):1-12.
[15] ZOU B C, QIU H D, LU Y F. Point cloud reduction and denoising based on optimized downsampling and bilateral filtering[J]. IEEE Access, 2020, 8:136316-136326.
[16] CHARLES R Q, HAO S, MO K C, et al. PointNet:deep learning on point sets for 3D classification and segmentation[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Washington D.C.,USA:IEEE Press,2017:652-660.
[17] THOMAS H, QI C R, DESCHAUD J E, et al. KPConv:flexible and deformable convolution for point clouds[C]//Proceedings of IEEE/CVF International Conference on Computer Vision. Washington D.C.,USA:IEEE Press,2019:6410-6419.
[18] HU Q Y, YANG B, XIE L H, et al. RandLA-Net:efficient semantic segmentation of large-scale point clouds[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D.C.,USA:IEEE Press,2020:11108-11117.
[19] PISTILLI F, FRACASTORO G, VALSESIA D, et al. Learning robust graph-convolutional representations for point cloud denoising[J]. IEEE Journal of Selected Topics in Signal Processing, 2021, 15(2):402-414.
[20] AZZABOU N, PARAGIOS N, GUICHARD F. Image reconstruction using particle filters and multiple hypotheses testing[J]. IEEE Transactions on Image Processing:a Publication of the IEEE Signal Processing Society, 2010, 19(5):1181-1190.
[21] CAO J J, CHEN H, ZHANG J, et al. Normal estimation via shifted neighborhood for point cloud[J]. Journal of Computational and Applied Mathematics, 2018, 329(C):57-67.
[22] ZHANG J, CAO J J, LIU X P, et al. Multi-normal estimation via pair consistency voting[J]. IEEE Transactions on Visualization and Computer Graphics, 2019, 25(4):1693-1706.
[23] 李军侠,王星驰,殷梓,等.边缘深度挖掘的弱监督显著性目标检测[J].计算机工程, 2023, 49(7):169-178. LI J X, WANG X C, YIN Z, et al. Weakly supervised salient object detection via edge depth mining[J]. Computer Engineering, 2023, 49(7):169-178.(in Chinese)
[24] 林洪彬,刘彬,张玉存.基于多尺度张量分解的点云结构特征提取[J].中国机械工程, 2012, 23(15):1833-1839. LIN H B, LIU B, ZHANG Y C. Structural feature extraction from point clouds based on multi-scale tensor decomposition[J]. China Mechanical Engineering, 2012, 23(15):1833-1839.(in Chinese)
[25] LIU Z, XIAO X W, ZHONG S S, et al. A feature-preserving framework for point cloud denoising[J]. Computer-Aided Design, 2020, 127:102857.
[26] ZHENG Y L, LI G Q, WU S H, et al. Guided point cloud denoising via sharp feature skeletons[J]. The Visual Computer, 2017, 33(6):857-867.
[27] JENKE P, WAND M, BOKELOH M, et al. Bayesian point cloud reconstruction[J]. Computer Graphics Forum, 2006, 25(3):379-388.
[28] 苏富林,钱素娟,魏霖静,等.频域内基于邻域特征学习的单幅图像超分辨重建[J].计算机工程, 2017, 43(5):255-260. SU F L, QIAN S J, WEI L J, et al. Single-image super-resolution reconstruction in frequency domain based on neighborhood feature learning[J]. Computer Engineering, 2017, 43(5):255-260.(in Chinese)
[29] 胡志隆,文畅,谢凯,等.联合HMM-UBM与RVM的声纹密码识别算法[J].计算机工程, 2018, 44(11):129-134. HU Z L, WEN C, XIE K, et al. Voiceprint password recognition algorithm fusing with HMM-UBM and RVM[J]. Computer Engineering, 2018, 44(11):129-134.(in Chinese)
[30] YADAV S K, REITEBUCH U, SKRODZKI M, et al. Constraint-based point set denoising using normal voting tensor and restricted quadratic error metrics[J]. Computers&Graphics, 2018, 74:234-243.
[31] SUN X F, ROSIN P, MARTIN R, et al. Fast and effective feature-preserving mesh denoising[J]. IEEE Transactions on Visualization and Computer Graphics, 2007, 13(5):925-938.

选择文件类型/文献管理软件名称

选择包含的内容