基于边缘区域特征细化的无监督点云补全网络研究

doi:10.3778/j.issn.1002-8331.2407-0232

摘要/Abstract

摘要： 现有点云补全方法通常利用区域特征捕捉不同区域的几何结构，而忽略了点云的边缘特征。这些特征描述了点云拐角处或复杂边界处的细节，是恢复精细点云的关键。为了实现区域特征和边缘细节的优势互补，提出了一种基于边缘区域特征细化的无监督点云补全网络。设计区域-边缘特征提取器，引入DGCNN（dynamic graph convolutional neural network）捕获点云的区域特征，并通过构建点云空间位置的归一化图筛选出边缘特征。提出缺失点生成器，将所提取的边缘特征融入局部区域，生成一个相对完整的粗糙点云。构建无监督点细化器，预测出描述点云结构的关键点，并进一步将粗糙点云细化为具有丰富细节的完整点云。在PCN、ShapeNet-55/34、MVP、KITTI数据集上的大量实验证明，该方法获得了先进的补全性能，相比于最新方法SeedFormer和ProxyFormer，MMD指数性能上分别提升了3.7%、2.2%。

关键词: 点云补全, 边缘特征, 特征互补, 无监督细化器

Abstract: Existing point cloud completion methods focus on area features for geometric structure capture, but often overlook critical edge details, which describe the details at the corners or complex boundaries of the point cloud, and are the key to recover the refined point cloud. In order to achieve the complementary advantages of region features and edge details, an unsupervised point cloud completion network based on edge regional feature refinement is proposed. A region-edge feature extractor is designed, which captures the region features of the point cloud via DGCNN (dynamic graph convolutional neural network), and then filters the edge features by constructing a normalized map of the spatial location of the point cloud. A missing-point generator is proposed to incorporate the extracted edge features into the local region to generate a relatively complete rough point cloud. An unsupervised point refiner is constructed to predict key points describing the structure, and further refines the rough point cloud into a complete point cloud with rich details. Extensive experiments on PCN, ShapeNet-55/34, MVP and KITTI datasets demonstrate significant performance gains. In particular, the MMD index is improved by 3.7% and 2.2% compared to the SeedFormer and ProxyFormer.

Key words: point cloud completion, edge features, feature complementarity, unsupervised point refiner

陈汉秋, 张惊雷, 贾鑫. 基于边缘区域特征细化的无监督点云补全网络研究[J]. 计算机工程与应用, 2025, 61(10): 192-202.

CHEN Hanqiu, ZHANG Jinglei, JIA Xin. Research on Unsupervised Point Cloud Completion Network Based on Edge Regional Feature Refinement[J]. Computer Engineering and Applications, 2025, 61(10): 192-202.

参考文献

[1] 熊珍凯, 程晓强, 吴幼冬, 等. 基于激光雷达的无人驾驶3D多目标跟踪[J]. 自动化学报, 2023, 49(10): 2073-2083.
XIONG Z K, CHENG X Q, WU Y D, et al. LiDAR-based 3D multi-object tracking for unmanned vehicles[J]. Acta Automatica Sinica, 2023, 49(10): 2073-2083.
[2] 白静, 徐昊. 多层次感知的多视图三维模型重建[J]. 计算机工程与应用, 2023, 59(2): 232-239.
BAI J, XU H. Multi-view 3D model reconstruction based on multi-level perception[J]. Computer Engineering and Applications, 2023, 59(2): 232-239.
[3] 赵阳, 李俊诚, 成博栋, 等. 深度学习在口腔医学影像中的应用与挑战[J]. 中国图象图形学报, 2024, 29(3): 586-607.
ZHAO Y, LI J C, CHENG B D, et al. Applications and challenges of deep learning in dental imaging[J]. Journal of Image and Graphics, 2024, 29(3): 586-607.
[4] WEN X, LI T Y, HAN Z Z, et al. Point cloud completion by skip?attention network with hierarchical folding[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 1936-1945.
[5] FEI B, YANG W D, CHEN W M, et al. Comprehensive review of deep learning-based 3D point cloud completion processing and analysis[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(12): 22862-22883.
[6] WEN X, XIANG P, HAN Z Z, et al. PMP-net: point cloud completion by learning multi-step point moving paths[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 7439-7448.
[7] YAN X J, YAN H Y, WANG J J, et al. FBNet: feedback network for point cloud completion[C]//Proceedings of the 17th European Conference on Computer Vision. Cham: Springer, 2022: 676-693.
[8] WANG Y D, TAN D J, NAVAB N, et al. Learning local displacements for point cloud completion[C]//Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 1558-1567.
[9] WU H, MIAO Y B. LRA-Net: local region attention network for 3D point cloud completion[C]//Proceedings of the 13th International Conference on Machine Vision, 2021: 357-367.
[10] YUAN W T, KHOT T, HELD D, et al. PCN: point completion network[C]//Proceedings of the 2018 International Conference on 3D Vision. Piscataway: IEEE, 2018: 728-737.
[11] CHARLES R Q, HAO S, MO K C, et al. PointNet: deep learning on point sets for 3D classification and segmentation[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 77-85.
[12] QI C R, YI L, SU H, et al. PointNet++: deep hierarchical feature learning on point sets in a metric space[C]//Advances in Neural Information Processing Systems 30, 2017: 5105-5114.
[13] XIANG P, WEN X, LIU Y S, et al. SnowflakeNet: point cloud completion by snowflake point deconvolution with skip-transformer[C]//Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 5479-5489.
[14] ZHOU H R, CAO Y, CHU W Q, et al. SeedFormer: patch seeds based point cloud completion with upsample transformer[C]//Proceedings of the 17th European Conference on Computer Vision. Cham: Springer, 2022: 416-432.
[15] HUANG X, HAN B, NING Y Q, et al. Edge-based feature extraction module for 3D point cloud shape classification[J]. Computers & Graphics, 2023, 112: 31-39.
[16] WANG Y, SUN Y B, LIU Z W, et al. Dynamic graph CNN for learning on point clouds[J]. ACM Transactions on Graphics, 2019, 38(5): 1-12.
[17] GOODFELLOW I, BENGIO Y, COURVILLE A. Deep learning[M]. Cambridge: MIT Press, 2016.
[18] HUANG Z T, YU Y K, XU J W, et al. PF-Net: point fractal network for 3D point cloud completion[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 7659-7667.
[19] XIE H Z, YAO H X, ZHOU S C, et al. GRNet: gridding residual network for dense point cloud completion[C]//Proceedings of the 16th European Conference on Computer Vision. Cham: Springer, 2020: 365-381.
[20] YU X M, RAO Y M, WANG Z Y, et al. PoinTr: diverse point cloud completion with geometry-aware transformers[C]//Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 12478-12487.
[21] CHEN Z K, LONG F C, QIU Z F, et al. AnchorFormer: point cloud completion from discriminative nodes[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 13581-13590.
[22] LI S S, GAO P, TAN X Y, et al. ProxyFormer: proxy alignment assisted point cloud completion with missing part sensitive transformer[C]//Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 9466-9475.
[23] 陆春媚, 杨志景. 多级精细化反卷积点云补全网络[J]. 计算机工程与应用, 2023, 59(17): 242-249.
LU C M, YANG Z J. Multistage refinement of deconvolution point cloud complementation network[J]. Computer Engineering and Applications, 2023, 59(17): 242-249.
[24] YANG Y Q, FENG C, SHEN Y R, et al. FoldingNet: point cloud auto-encoder via deep grid deformation[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 206-215.
[25] DUAN C, CHEN S, TIAN D, et al. Deep graph topology learning for 3D point cloud reconstruction: TR2019-046 [R]. Mitsubishi Electric Research Laboratories, 2019.
[26] GROUEIX T, FISHER M, KIM V G, et al. A papier- maché approach to learning 3D surface generation[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 216-224.
[27] TCHAPMI L P, KOSARAJU V, REZATOFIGHI H, et al. TopNet: structural point cloud decoder[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 383-392.
[28] WANG J, CUI Y, GUO D Y, et al. PointAttN: you only need attention for point cloud completion[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(6): 5472-5480.
[29] TANG J S, GONG Z J, YI R, et al. LAKe-net: topology-aware point cloud completion by localizing aligned keypoints[C]//Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 1716-1725.
[30] RONG Y, ZHOU H R, YUAN L X, et al. CRA-PCN: point cloud completion with intra- and inter-level cross-resolution transformers[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(5): 4676-4685.
[31] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Advances in Neural Information Processing Systems 30, 2017: 6000-6010.
[32] GEIGER A, LENZ P, STILLER C, et al. Vision meets robotics: the KITTI dataset[J]. International Journal of Robotics Research, 2013, 32(11): 1231-1237.
[33] PAN L, CHEN X, CAI Z, et al. Variational relational point completion network[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 8524-8533.
[34] WANG X G, ANG M H, LEE G H. Cascaded refinement network for point cloud completion[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 787-796.
[35] LIU M H, SHENG L, YANG S, et al. Morphing and sampling network for dense point cloud completion[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(7): 11596-11603.