结合有序建模和空间感知的点云语义分割算法

Ordered Modeling and Spatial Perception for Point Cloud Semantic Segmentation Algorithm

摘要

针对三维点云语义分割中的无序性、几何信息缺失及边界模糊等挑战，提出一种结合有序建模和空间感知的点云语义分割算法。首先通过融合 Z 阶与希尔伯特空间填充曲线及其镜像变体进行多种空间填充曲线的互补有序建模，构建空间结构保持性互补的点云编码序列；然后设计动态空间编码机制，根据不同阶段任务需求自适应调整编码维度，提升空间编码效率；最后构建空间感知聚合模块，融合空间引导的局部特征传播与多层感知机全局上下文感知，实现特征高效混合学习，增强特征的空间稳定性与几何一致性。在 S3DIS、ScanNet v2 和 ScanObjectNN 3 个公开点云数据集上的实验表明，所提算法实现了高精度语义分割，有效提升了对复杂边界区域和小物体的空间感知能力和语义理解精度。

Abstract

To address the challenges of disorder，missing geometric information，and blurred boundaries in 3D point cloud semantic segmentation，this paper proposes an algorithm that combines complementary ordered modeling with spatial perception.A complemen- tary ordered modeling module is first constructed by integrating Z-order curves，Hilbert curves，and their mirrored variants，forming spatially consistent and structurally complementary point cloud encoding sequences.Then，a dynamic spatial encoding mechanism is in- troduced to adaptively adjust the encoding dimensions according to the requirements of different stages，thereby improving the efficien- cy of spatial representation.Finally，a spatial-aware aggregation module is designed to fuse spatially guided local feature propagation with multilayer perceptron-based global context modeling，achieving efficient hybrid feature learning while enhancing spatial stability and geometric consistency.Experiments conducted on three public datasets-S3DIS，ScanNet v2，and ScanObjectNN-demonstrate that the proposed algorithm delivers high-accuracy semantic segmentation and significantly enhances spatial perception and semantic under- standing in complex boundary regions and small object scenarios.

关键词

三维点云语义分割 / 空间填充曲线 / 互补有序建模 / 动态空间编码 / 空间感知聚合

Key words

3D point cloud semantic segmentation / space-filling curves / complementary ordered modeling / dynamic spatial encoding / spatial-aware aggregation

参考文献

[1] Qu W, Wang J, Gong Y, et al. An end-to-end robust point cloud se- mantic segmentation network with single-step conditional diffusion models[C]// Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, 2025:27325-27335.

[2] Zhang J, Lin X, et al. SVM-based classification of segmented air- borne LiDAR point clouds in urban areas[J]. Remote Sensing, 2013, 5(8):3749-3775.

[3] Qi C R, Su H, Mo K, et al. PointNet :deep learning on point sets for 3d classification and segmentation[C]// Proceedings of the IEEE/ CVF Conference on Computer Vision and Pattern Recognition, 2017:652-660.

[4] Zhao H, Jiang L, Jia J, et al. Point transformer[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021:16259-16268.

[5] Peano G. Sur une courbe,qui remplit toute une aireplane[M]. Vi-enna: Springer Vienna,1990:71-75.

[6] Chen X, Wu Y, Xu W, et al. Pointscnet:point cloud structure and correlation learning based on space filling curve-guided sampling[J]. Symmetry, 2021, 14(1):8,doi:https://doi.org/10.3390/sym14010008.

[7] Qi C R, Yi L, Su H, et al. Pointnet＋＋:deep hierarchical feature learning on point sets in a metric space[C]// Proceedings of the 31st International Conference on Neural Information Processing Systems, 2017:5105-5114.

[8] Qian G, Li Y, Peng H, et al. Pointnext:revisiting pointnet＋＋with improved training and scaling strategies[C]// Proceedings of the 36th International Conference on Neural Information Processing Systems, 2022:23192-23204.

[9] Chen X, Ma II, Wan J, et al. Multi-view 3D object detection net- work for autonomous driving[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2017:1907-1915.

[10] Xu C, Wu B, Wang Z, et al. Squeezesegv3 :spatially-adaptive con- volution for efficient point-cloud segmentation[C]// Proceedings of the European Conference on Computer Vision, 2020:1-19.

[11] Wu Z, Song S, Khosla A, et al. 3D shapenets :a deep representation for volumetric shapes[C]// Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, 2015:1912-1920.

[12] Chen Y, Liu J, Zhang X, et al. Voxelnext:fully sparse voxelnet for 3d object detection and tracking[C]// Proceedings of the IEEE/ CVF Conference on Computer Vision and Pattern Recognition, 2023:21674-21683.

[13] LU C J, PAN X. Multiscale dynamic graph CNN with edge supervi- sion loss for point cloud segmentation[J]. Journal of Chinese Computer Systems, 2024, 45(9):2215-2220.

[14] Lai X, Liu J, Jiang L, et al. Stratified transformer for 3D point cloud segmentation[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022:8500-8509.

[15] Wu X, Jiang L, Wang P S, et al. Point transformer v3 :simpler faster stronger[C]// Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, 2024:4840-4851.

[16] Liang D, Zhou X, Xu W, et al. Pointmamba:a simple state space model for point cloud analysis[C]// Proceedings of the 38th Inter- national Conference on Neural Information Processing Systems, 2024:32653-32677.

[17] Han X, Tang Y, Wang Z, et al. Mamba3d:enhancing local features for 3D point cloud analysis via state space model[ C]// Proceed- ings of the 32nd ACM International Conference on Multimedia, 2024:4995-5004.

[18] Wang P S. OctFormer:octree-based transformers for 3D point clouds[J]. ACM Transactions on Graphics, 2023, 42(4):1-11.

[19] Zhang T, Yuan H, Qi L, et al. Point cloud mamba:point cloud learning via state space model[ C]// Proceedings of the AAAI Conference on Artificial Intelligence, 2025;10121 10130.

[20] Li Z, Ai Y, Lu J, et al. Pamba:enhancing global interaction in point clouds via state space model[C]// Proceedings of the AAAI Con- ference on Artificial Intelligence, 2025:5092-5100.

[21] Li G, Müller M, Thabet A, et al. DeepGCNs: can GCNs go as deep as CNNs?[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019:9267-9276.

[22] Yu W, Luo M, Zhou P, et al. Metaformer is actually what you need for vision[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022:10819-10829.

[23] Wu X, Lao Y, Jiang L, et al. Point transformer v2: grouped vector attention and partition-based pooling[C]// Proceedings of the 36th International Conference on Neural Information Processing Sys- tems, 2022:33330-33342.

[24] Chen B, Xia Y, Zang Y, et al. Decoupled local aggregation for point cloud learning[EB/OL]. https://arxiv.org/pdf/2308.16532,2023.

[25] Zeng Z, Dong M, Zhou J, et al. Deepla-Net:very deep local aggre- gation networks for point cloud analysis[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion, 2025:1330-1341.

[26] 卢超杰, 潘翔. 逃缘损失约束下的多尺度动态图卷积点云分割网络[J]. 小型微型计算机系统, 2024, 45(9):2215-2220.

基金资助

国家自然科学基金项目(62262036)

国家自然科学基金项目(62362043)

云南省兴滇英才支持计划项目(KKXY202203008)