联合边界引导和多尺度特征的路侧交通地物点云感知

Roadside traffic object point cloud perception with joint boundary guidance and multi-scale features

摘要

针对复杂城市场景中路侧交通地物形态多样、边界模糊且易受遮挡，导致点云感知精度不稳定的问题，提出了一种面向路侧交通地物精确感知的新方法。该方法通过在点云特征学习过程中引入边界约束，减少不同地物之间的相互干扰，并结合多尺度信息融合策略，同时兼顾局部细节与整体结构，从而提升在复杂道路环境中的识别能力。实验采用 3 份不同车载激光扫描系统获取的不同城市场景数据来验证本方法的有效性。结果表明，所提模型在 3 组数据上的平均交并比和路侧交通地物交并比分别达到 89.48% 和 87.91%，并在快速路、主干路、次干路和支路等不同道路类型中分别取得 94.90%、93.36%、90.14% 和 88.56% 的路侧交通地物交并比。该方法在多种道路场景下均能够稳定提升路侧交通地物的识别精度，尤其在地物密集、遮挡严重的城市街道环境中表现出更强的鲁棒性与泛化能力。研究结果可为推动交通基础设施数字化转型提供数据基础。

Abstract

To address the problem of unstable point cloud perception accuracy caused by the diverse forms, blurred boundaries and high susceptibility to occlusions of roadside traffic objects in complex urban environments, this paper proposes a novel method for precise perception of roadside traffic objects. By introducing boundary constraints into the point cloud feature learning process, the proposed method effectively mitigates interference among different object categories. Meanwhile, a multi-scale feature fusion strategy is incorporated to jointly capture local details and global structural information, thereby enhancing recognition performance in complex road scenarios. Experiments are conducted on three datasets collected from different urban scenes using vehicle-mounted LiDAR systems to validate the effectiveness of the proposed approach. The results demonstrate that the proposed model achieves a mean Intersection over Union (mIoU) of 89.48% and a roadside traffic object IoU of 87.91% across the three datasets. Furthermore, the method attains IoU scores of 94.90%, 93.36%, 90.14%, and 88.56% for roadside traffic objects on expressways, arterial roads, secondary trunk roads, and branch roads, respectively. Overall, the proposed method consistently improves the recognition accuracy of roadside traffic objects across diverse road environments, exhibiting strong robustness and generalization capability, particularly in densely populated and heavily occluded urban street scenes. The findings of this study provide a reliable data foundation for advancing the digital transformation of transportation infrastructure.

关键词

路侧交通地物 / 车载激光扫描系统 / 边界引导 / 多尺度特征融合 / 鲁棒性

Key words

roadside traffic objects / mobile laser scanning systems / boundary guidance / multi-scale feature fusion / robustness

作者贡献声明

赖启印负责论文的构思与设计，数据收集与分析处理，论文的撰写。方莉娜提供论文经费与资源支持，指导与监督，负责论文的修改。

AI使用声明

本文在完成初稿后使用AI对文字内容进行了润色，并在此基础上进行了修改。

利益冲突声明

不存在已知的可能影响本论文所报告工作的竞争性经济利益或个人关系。

伦理声明

本文不涉及人类或动物的生物学研究。

数据可用性

实验数据Ⅱ与实验数据Ⅲ为公开数据集，如需获取与本文相关的其他数据，可向通信作者提出申请。

参考文献

[1] 杨必胜， 陈驰， 董震 . 面向智能化测绘的城市地物三维提取 [J]. 测绘学报， 2022， 51(7)： 1476-1484.

[2] Yang B S， Chen C， Dong Z . 3D geospatial information extraction of urban objects for smart surveying and mapping[J]. Acta Geodaetica et Cartographica Sinica， 2022， 51(7)： 1476-1484.

[3] 郭仁忠， 陈业滨， 赵志刚， 等 . GIS 的科学概念转化：从 Map—based GIS 到 Space—oriented GIS[J]. 测绘学报， 2024， 53(10)： 1853-1862.

[4] Guo R Z， Chen Y B， Zhao Z G， et al. The transformation of the scientific concept of GIS：from Map—based GIS to Space—oriented GIS[J]. Acta Geodaetica et Cartographica Sinica， 2024， 53(10)： 1853-1862.

[5] Riveiro B， González—jorge H， Martínez—sánchez J， et al. Automatic detection of zebra crossings from mobile LiDAR data[J]. Optics & Laser Technology， 2015， 70： 63-70.

[6] 刘华 . 车载激光点云地物提取与分类研究 [J]. 测绘学报， 2020， 49(11)： 1506.

[7] Liu H. Object points extraction and classification of mobile LiDAR point clouds[J]. Acta Geodaetica et Cartographica Sinica， 2020， 49(11)： 1506.

[8] 米晓新， 杨必胜， 董震 . 车载激光点云道路场景可视域快速计算与应用 [J]. 武汉大学学报（信息科学版）， 2020， 45(2)： 258-264.

[9] Mi X X， Yang B S， Dong Z . Fast visibility analysis and application in road environment with mobile laser scanning data[J]. Geomatics and Information Science of Wuhan University， 2020， 45(2)： 258-264.

[10] 杨必胜， 董震 . 点云智能研究进展与趋势 [J]. 测绘学报， 2019， 48(12)： 1575-1585.

[11] Yang B S， Dong Z . Progress and perspective of point cloud intelligence[J]. Acta Geodaetica et Cartographica Sinica， 2019， 48(12)： 1575-1585.

[12] Xu Y S， Sun Z H， Hoegner L， et al. Instance segmentation of trees in urban areas from MLS point clouds using supervoxel contexts and graph—based optimization[C]// Proceedings of the 2018 10th IAPR Workshop on Pattern Recognition in Remote Sensing. Beijing：IEEE， 2018： 1-5.

[13] 董震， 杨必胜 . 车载激光扫描数据中多类目标的层次化提取方法 [J]. 测绘学报， 2015， 44(9)： 980-987.

[14] Dong Z， Yang B S . Hierarchical extraction of multiple objects from mobile laser scanning data[J]. Acta Geodaetica et Cartographica Sinica， 2015， 44(9)： 980-987.

[15] Xu M B， Zhong X T， Zhong R F . A method for automatic extraction and individual segmentation of urban street trees from laser point clouds[J]. Optics & Laser Technology， 2025， 180： 111431.

[16] Fan W， Chenglu W， Jonathan L . Automated extraction of urban trees from mobile LiDAR point clouds[C]// Proceedings of SPIE 9901，2nd ISPRS International Conference on Computer Vision in Remote Sensing. Xiamen：SPIE， 2016： 159-164.

[17] 王鹏， 刘如飞， 马新江 . 车载激光点云中杆状目标自动提取 [J]. 遥感信息， 2017， 32(5)： 52-56.

[18] Wang P， Liu R F， Ma X J . Automatic extraction of pole—like objects from vehicle borne laser scanning point clouds[J]. Remote Sensing Information， 2017， 32(5)： 52-56.

[19] Yang B S， Dong Z， Liu Y， et al. Computing multiple aggregation levels and contextual features for road facilities recognition using mobile laser scanning data[J]. ISPRS Journal of Photogrammetry and Remote Sensing， 2017， 126： 180-194.

[20] Lehtomäki M， Jaakkola A， Hyyppä J， et al. Object classification and recognition from mobile laser scanning point clouds in a road environment[J]. IEEE Transactions on Geoscience and Remote Sensing， 2016， 54(2)： 1226-1239.

[21] 李永强， 李鹏鹏， 董亚涵， 等 . 车载 LiDAR 点云数据中杆状地物自动提取与分类 [J]. 测绘学报， 2020， 49(6)： 724-735.

[22] Li Y Q， Li P P， Dong Y H， et al. Automatic extraction and classification of pole—like objects from vehicle LiDAR point cloud[J]. Acta Geodaetica et Cartographica Sinica， 2020， 49(6)： 724-735.

[23] Yu Y T， Li J， Yu J， et al. Pairwise three—dimensional shape context for partial object matching and retrieval on mobile laser scanning data[J]. IEEE Geoscience and Remote Sensing Letters， 2014， 11(5)： 1019-1023.

[24] Qi C R， Su H， Mo K， et al. Pointnet：deep learning on point sets for 3d classification and segmentation[C]// Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu：IEEE， 2017： 77-85.

[25] Wang Y， Sun Y B， Liu Z W， et al. Dynamic graph CNN for learning on point clouds[J]. ACM Transactions on Graphics（TOG）， 2019， 38(5)： 146.

[26] Thomas H， Qi C R， Deschaud J E， et al. KPconv：flexible and deformable convolution for point clouds[C]// Proceedings of 2019 IEEE/CVF International Conference on Computer Vision. Seoul：IEEE， 2019： 6410-6419.

[27] Hu Q Y， Yang B， Xie L H， et al. RandLA—Net：efficient semantic segmentation of large—scale point clouds[C]// Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle：IEEE， 2020： 11105-11114.

[28] Fan S Q， Dong Q L， Zhu F H， et al. SCF—Net：learning spatial contextual features for large—scale point cloud segmentation[C]// Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville：IEEE， 2021： 14499-14508.

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

[30] 汪小钦， 王苗苗， 王绍强， 等 . 基于可见光波段无人机遥感的植被信息提取 [J]. 农业工程学报， 2015， 31(5)： 152-159.

[31] Wang X Q， Wang M M， Wang S Q， et al. Extraction of vegetation information from visible unmanned aerial vehicle images[J]. Transactions of the Chinese Society of Agricultural Engineering， 2015， 31(5)： 152-159.

[32] Chen J Y， Fang Q， Zhang D L， et al. A critical review of automated extraction of rock mass parameters using 3D point cloud data[J]. Intelligent Transportation Infrastructure， 2023， 2： liad005.

[33] Gong J Y， Xu J C， Tan X， et al. Boundary—aware geometric encoding for semantic segmentation of point clouds[C]// Proceedings of the 35th AAAI Conference on Artificial Intelligence. [S. l.]：AAAI， 2021： 11423-11432.

[34] Koonce B. ResNet 50[M]//KOONCE B. Convolutional neural networks with swift for tensorflow：image recognition and dataset categorization. Berkeley：Apress， 2021： 63-72.

[35] Mao A Q， Mohri M， Zhong Y T . Cross—entropy loss functions：theoretical analysis and applications[C]// Proceedings of the 40th International Conference on Machine Learning. Hawaii：PMLR， 2023： 23803-23828.

[36] Cui Y， Jia M L， Lin T Y， et al. Class—balanced loss based on effective number of samples[C]// Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach：IEEE， 2019： 9260-9269.

[37] Tang L Y， Zhan Y B， Chen Z， et al. Contrastive boundary learning for point cloud segmentation[C]// Proceedings of 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans：IEEE， 2022： 8479-8489.

[38] Tan W K， Qin N N， Ma L F， et al. Toronto—3D：a large—scale mobile LiDAR dataset for semantic segmentation of urban roadways[C]// Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Seattle：IEEE， 2020： 797-806.

[39] Roynard X， Deschaud J E， Goulette F . Paris—Lille—3D：a large and high—quality ground—truth urban point cloud dataset for automatic segmentation and classification[J]. The International Journal of Robotics Research， 2018， 37(6)： 545-557.

[40] Shuai H， Xu X， Liu Q S， et al. Backward attentive fusing network with local aggregation classifier for 3D point cloud semantic segmentation[J]. IEEE Transactions on Image Processing， 2021， 30： 4973-4984.