基于点云指纹的定位与建图方法

doi:10.16183/j.cnki.jsjtu.2024.430

上海交通大学学报 ›› 2026, Vol. 60 ›› Issue (8): 1374-1384.doi: 10.16183/j.cnki.jsjtu.2024.430

基于点云指纹的定位与建图方法

李飞¹^,², 刘小汇³(), 陈佳良²^,⁴, 袁粤林¹^,², 文超³

¹ 北京航空航天大学空间与环境学院, 北京 100191
² 国防科技大学前沿交叉学科学院, 长沙 410073
³ 国防科技大学电子科学学院, 长沙 410073
⁴ 大连理工大学建设工程学院, 辽宁大连 116024

收稿日期:2024-10-28 修回日期:2025-01-18 接受日期:2025-03-07 出版日期:2026-08-28 发布日期:2026-09-02
通讯作者: 刘小汇,研究员,博士生导师;E-mail:liuxh@nudt.edu.cn.
作者简介:李飞(1993—),硕士,科研工程师,从事智能体高精定位研究.
基金资助:
国家自然科学基金(U20A20193)

Point Cloud Fingerprint-Based Localization and Mapping

LI Fei¹^,², LIU Xiaohui³(), CHEN Jialiang²^,⁴, YUAN Yuelin¹^,², WEN Chao³

¹ School of Space and Environment, Beihang University, Beijing 100191, China
² College of Advanced Interdisciplinary Studies, National University of Defense Technology, Changsha 410073, China
³ College of Electronic Science and Technology, National University of Defense Technology, Changsha 410073, China
⁴ School of Infrastructure Engineering, Dalian University of Technology, Dalian 116024, Liaoning, China

Received:2024-10-28 Revised:2025-01-18 Accepted:2025-03-07 Online:2026-08-28 Published:2026-09-02

摘要/Abstract

摘要：

为提升基于激光雷达的智能车定位性能,引入激光点云指纹的概念,并提出基于点云指纹的地图表征模型与智能车定位方法.基于点云极化与主成分特征的点云指纹表征融合了点云的全局与局部描述.地图表征模型由序列节点构成,每个节点仅包含激光点云指纹与全局位姿信息.将基于点云指纹地图的智能车多尺度定位解耦成3个模块:基于普通全球导航卫星系统或轨迹预测的粗定位;利用点云指纹与基于向量的皮尔逊相关系数的节点定位;基于点云指纹与快速广义迭代最近点的度量定位.基于公开的KITTI数据集与本地数据集的实验结果表明,所提方法取得超过99%的地图匹配精度与高达7 cm的定位精度,且在不同场景下均取得厘米级高精定位,鲁棒性强且泛化性好.

关键词: 点云指纹, 极化表征, 主成分特征, 地图表征, 多尺度定位

Abstract:

To enhance the localization performance of intelligent vehicles using light detection and ranging (LiDAR), the concept of point cloud fingerprints is introduced, along with a map representation model and a localization method based on point cloud fingerprints. The point cloud fingerprint representation is conducted based on point cloud polarization and principal component features, integrating global and local descriptions of the point cloud. The map representation model consists of sequential nodes containing only point cloud fingerprints and global poses. Multiscale localization of intelligent vehicles based on point cloud fingerprint maps is decoupled into three modules: coarse localization based on standard global navigation satellite system (GNNS) data or trajectory prediction, node localization based on point cloud fingerprints and vector-based Pearson correlation coefficient, and metric localization based on point cloud fingerprints and fast generalized iterative closest point. Experiments on the public KITTI dataset and a local dataset demonstrate that the proposed method achieves over 99% map-matching accuracy and up to 7 cm localization accuracy. The results indicate that the proposed method attains centimeter-level high-precision localization in various scenarios with strong robustness and generalization capability.

Key words: point cloud fingerprint, polarization representation, principal component feature, map representation, multiscale localization

中图分类号:

TP242.6

李飞, 刘小汇, 陈佳良, 袁粤林, 文超. 基于点云指纹的定位与建图方法[J]. 上海交通大学学报, 2026, 60(8): 1374-1384.

LI Fei, LIU Xiaohui, CHEN Jialiang, YUAN Yuelin, WEN Chao. Point Cloud Fingerprint-Based Localization and Mapping[J]. Journal of Shanghai Jiao Tong University, 2026, 60(8): 1374-1384.

图/表 12

图1

图2

图3

表1

图4

表2

表3

图5

表4

表5

表6

图6

参考文献 25

[1]	HUANG L Y. Review on LiDAR-based SLAM techniques[C]// 2021 International Conference on Signal Processing & Machine Learning. Stanford, USA: IEEE, 2021: 163-168.
[2]	ZHANG J, SINGH S. Low-drift and real-time lidar odometry and mapping[J]. Autonomous Robots, 2017, 41(2): 401-416. doi: 10.1007/s10514-016-9548-2 URL
[3]	SHAN T X, ENGLOT B. LeGO-LOAM: Lightweight and ground-optimized lidar odometry and mapping on variable terrain[C]// 2018 IEEE/RSJ International Conference on Intelligent Robots & Systems. Madrid, Spain: IEEE, 2018: 4758-4765.
[4]	WANG H, WANG C, CHEN C L, et al. F-LOAM: Fast LiDAR odometry and mapping[C]// 2021 IEEE/RSJ International Conference on Intelligent Robots & Systems. Prague, Czech Republic: IEEE, 2021: 4390-4396.
[5]	STILLA U, XU Y S. Change detection of urban objects using 3D point clouds: A review[J]. ISPRS Journal of Photogrammetry & Remote Sensing, 2023, 197: 228-255.
[6]	HE L, WANG X L, ZHANG H. M2DP: A novel 3D point cloud descriptor and its application in loop closure detection[C]// 2016 IEEE/RSJ International Conference on Intelligent Robots & Systems. Daejeon, ROK: IEEE, 2016: 231-237.
[7]	KIM G, KIM A. Scan context: Egocentric spatial descriptor for place recognition within 3D point cloud map[C]// 2018 IEEE/RSJ International Conference on Intelligent Robots & Systems. Madrid, Spain: IEEE, 2018: 4802-4809.
[8]	SIDOR K, WYSOCKI M. Recognition of human activities using depth maps and the viewpoint feature histogram descriptor[J]. Sensors, 2020, 20(10): 2940. doi: 10.3390/s20102940 URL
[9]	ZHOU Z C, ZHAO C, ADOLFSSON D, et al. NDT-transformer: Large-scale 3D point cloud localisation using the normal distribution transform representation[C]// 2021 IEEE International Conference on Robotics & Automation. Xi’an, China: IEEE, 2021: 5654-5660.
[10]	CATTANEO D, VAGHI M, VALADA A. LCDNet: Deep loop closure detection and point cloud registration for LiDAR SLAM[J]. IEEE Transactions on Robotics, 2022, 38(4): 2074-2093. doi: 10.1109/TRO.2022.3150683 URL
[11]	LI L, KONG X, ZHAO X R, et al. RINet: Efficient 3D lidar-based place recognition using rotation invariant neural network[J]. IEEE Robotics & Automation Letters, 2022, 7(2): 4321-4328.
[12]	MINH D, WANG H X, LI Y F, et al. Explainable artificial intelligence: A comprehensive review[J]. Artificial Intelligence Review, 2022, 55(5): 3503-3568. doi: 10.1007/s10462-021-10088-y
[13]	HAN X F, FENG Z A, SUN S J, et al. 3D point cloud descriptors: State-of-the-art[J]. Artificial Intelligence Review, 2023, 56(10): 12033-12083. doi: 10.1007/s10462-023-10486-4
[14]	LE CUN Y, KANTER I, SOLLA S A. Eigenvalues of covariance matrices: Application to neural-network learning[J]. Physical Review Letters, 1991, 66(18): 2396-2399. pmid: 10043474
[15]	ÖZBAY E, ÇINAR A. A comparative study of object classification methods using 3D zernike moment on 3D point clouds[J]. Traitement Du Signal, 2019, 36(6): 549-555. doi: 10.18280/ts URL
[16]	DUBÉ R, DUGAS D, STUMM E, et al. SegMatch: Segment based place recognition in 3D point clouds[C]// 2017 IEEE International Conference on Robotics & Automation. Singapore, Singapore: IEEE, 2017: 5266-5272.
[17]	DUBÉ R, CRAMARIUC A, DUGAS D, et al. SegMap: Segment-based mapping and localization using data-driven descriptors[J]. The International Journal of Robotics Research, 2020, 39(2/3): 339-355. doi: 10.1177/0278364919863090 URL
[18]	KONG L D, LIU Y Q, CHEN R N, et al. Rethinking range view representation for LiDAR segmentation[C]// 2023 IEEE/CVF International Conference on Computer Vision. Paris, France: IEEE, 2023: 228-240.
[19]	SMILDE A K, KIERS H A L, BIJLSMA S, et al. Matrix correlations for high-dimensional data: The modified RV-coefficient[J]. Bioinformatics, 2009, 25(3): 401-405. doi: 10.1093/bioinformatics/btn634 pmid: 19073588
[20]	KOIDE K, YOKOZUKA M, OISHI S, et al. Voxelized GICP for fast and accurate 3D point cloud registration[C]// 2021 IEEE International Conference on Robotics & Automation. Xi’an, China: IEEE, 2021: 11054-11059.
[21]	DENG L, LI G Q, HAN S, et al. Model compression and hardware acceleration for neural networks: A comprehensive survey[J]. Proceedings of the IEEE, 2020, 108(4): 485-532. doi: 10.1109/PROC.5 URL
[22]	WILDE D K. Computing clothoid segments for trajectory generation[C]// 2009 IEEE/RSJ International Conference on Intelligent Robots & Systems. St. Louis, USA: IEEE, 2009: 2440-2445.
[23]	GEIGER A, LENZ P, STILLER C, et al. Vision meets robotics: The KITTI dataset[J]. The International Journal of Robotics Research, 2013, 32(11): 1231-1237. doi: 10.1177/0278364913491297 URL
[24]	CUMMINS M, NEWMAN P. FAB-MAP: Probabilistic localization and mapping in the space of appearance[J]. The International Journal of Robotics Research, 2008, 27(6): 647-665. doi: 10.1177/0278364908090961 URL
[25]	DONG Z, LIANG F X, YANG B S, et al. Registration of large-scale terrestrial laser scanner point clouds: A review and benchmark[J]. ISPRS Journal of Photogrammetry & Remote Sensing, 2020, 163: 327-342.

数据	路径长度/m	是否存在闭环	场景类型	车速/ (km·h^-1)	持续时间/s
KITTI 00	3724	是	城区	28.41	472
KITTI 02	5063	是	城区	37.59	485
KITTI 05	2208	是	城区	27.60	288
KITTI 06	1234	是	城区	38.30	116
KITTI 07	693	是	城区	21.51	116
KITTI 09	1706	是	城区	37.00	166
本地数据	903	是	园区	10.12	321

参数	取值
θ^gd/(°)	10
δ/%	40
ϕ/%	95
$ \varsigma$/%	50
ε/m	1.0~2.5
ψ	5
N	50

LiDAR	原始数据大小/KB	极化矩阵		点云指纹
LiDAR	原始数据大小/KB	存储需求/KB	压缩比例/%	存储需求/KB	压缩比例/%
HDL-64E	1449.45	281.25	80.60	55.93	96.14
镭神C32	747.58	140.63	81.19	48.53	93.51

数据	ε/m	地图大小/MB	存储密度/ (MB·km^-1)
KITTI 00	≈ 1.5	124.28	33.37
KITTI 02	≈ 2.0	127.05	25.09
KITTI 05	≈ 1.5	75.51	34.20
KITTI 06	≈ 2.5	30.17	24.45
KITTI 07	≈ 1.0	30.14	43.49
KITTI 09	≈ 2.0	43.33	25.40
本地数据	≈ 1.0	39.64	43.90

数据	节点数	LeGO-LOAM的方法		Scan context		所提方法
数据	节点数	正确匹配数	准确率/%	正确匹配数	准确率/%	正确匹配数	准确率/%
KITTI 00	2270	1611	70.97	1935	85.29	2268	99.91
KITTI 02	2330	1605	68.88	1860	79.83	2292	98.37
KITTI 05	1380	983	71.23	1218	88.26	1376	99.71
KITTI 06	550	413	75.09	503	91.45	550	100.00
KITTI 07	550	399	72.55	461	83.82	547	99.45
KITTI 09	795	610	76.73	699	87.92	795	100.00
本地数据	1678	1194	71.16	1361	81.11	1662	99.05

基于点云指纹的定位与建图方法

Point Cloud Fingerprint-Based Localization and Mapping

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 25

相关文章 14

编辑推荐

Metrics

本文评价

类型	数据	定位精度/ m	最大误差/ m	单帧时间/ ms	内存占用/ %	类型
A-LOAM	KITTI 00	14.43	41.27	104.76	6.58	LeGO-LOAM	KITTI 07	0.43	1.12	104.48	5.12
A-LOAM	KITTI 02	44.04	122.48	103.36	6.09	LeGO-LOAM	KITTI 07	0.43	1.12	104.48	5.12
A-LOAM	KITTI 02	44.04	122.48	103.36	6.09	LeGO-LOAM	KITTI 09	9.90	25.57	104.31	5.61
A-LOAM	KITTI 05	7.57	20.21	102.52	6.83	LeGO-LOAM	KITTI 09	9.90	25.57	104.31	5.61
A-LOAM	KITTI 05	7.57	20.21	102.52	6.83	LeGO-LOAM	本地数据	7.50	12.65	44.80	5.12
A-LOAM	KITTI 06	6.00	17.08	106.06	5.36	LeGO-LOAM	本地数据	7.50	12.65	44.80	5.12
A-LOAM	KITTI 06	6.00	17.08	106.06	5.36	SC-LeGO-LOAM	KITTI 00	3.85	12.39	103.50	5.85
A-LOAM	KITTI 07	1.51	2.63	102.99	5.48	SC-LeGO-LOAM	KITTI 00	3.85	12.39	103.50	5.85
A-LOAM	KITTI 07	1.51	2.63	102.99	5.48	SC-LeGO-LOAM	KITTI 02	7.84	49.57	103.91	6.34
A-LOAM	KITTI 09	14.68	36.77	103.24	6.34	SC-LeGO-LOAM	KITTI 02	7.84	49.57	103.91	6.34
A-LOAM	KITTI 09	14.68	36.77	103.24	6.34	SC-LeGO-LOAM	KITTI 05	1.18	2.77	108.08	6.09
A-LOAM	本地数据	0.77	1.54	44.15	5.12	SC-LeGO-LOAM	KITTI 05	1.18	2.77	108.08	6.09
A-LOAM	本地数据	0.77	1.54	44.15	5.12	SC-LeGO-LOAM	KITTI 06	0.95	3.00	97.93	6.58
F-LOAM	KITTI 00	13.41	38.74	104.97	4.38	SC-LeGO-LOAM	KITTI 06	0.95	3.00	97.93	6.58
F-LOAM	KITTI 00	13.41	38.74	104.97	4.38	SC-LeGO-LOAM	KITTI 07	0.39	1.08	107.02	5.12
F-LOAM	KITTI 02	18.58	47.39	106.07	3.66	SC-LeGO-LOAM	KITTI 07	0.39	1.08	107.02	5.12
F-LOAM	KITTI 02	18.58	47.39	106.07	3.66	SC-LeGO-LOAM	KITTI 09	6.26	23.23	99.74	5.85
F-LOAM	KITTI 05	7.75	20.49	105.13	3.90	SC-LeGO-LOAM	KITTI 09	6.26	23.23	99.74	5.85
F-LOAM	KITTI 05	7.75	20.49	105.13	3.90	SC-LeGO-LOAM	本地数据	7.80	13.04	41.51	5.36
F-LOAM	KITTI 06	6.34	18.02	103.43	3.66	SC-LeGO-LOAM	本地数据	7.80	13.04	41.51	5.36
F-LOAM	KITTI 06	6.34	18.02	103.43	3.66	所提方法	KITTI 00	0.05	0.63	109.15	5.67
F-LOAM	KITTI 07	1.54	2.61	104.34	3.66	所提方法	KITTI 00	0.05	0.63	109.15	5.67
F-LOAM	KITTI 07	1.54	2.61	104.34	3.66	所提方法	KITTI 02	0.12	1.69	107.25	5.56
F-LOAM	KITTI 09	14.10	34.69	103.33	3.41	所提方法	KITTI 02	0.12	1.69	107.25	5.56
F-LOAM	KITTI 09	14.10	34.69	103.33	3.41	所提方法	KITTI 05	0.04	0.34	93.76	5.95
F-LOAM	本地数据	0.89	1.78	50.93	6.34	所提方法	KITTI 05	0.04	0.34	93.76	5.95
F-LOAM	本地数据	0.89	1.78	50.93	6.34	所提方法	KITTI 06	0.07	1.07	93.68	5.10
LeGO-LOAM	KITTI 00	4.43	11.81	102.50	5.12	所提方法	KITTI 06	0.07	1.07	93.68	5.10
LeGO-LOAM	KITTI 00	4.43	11.81	102.50	5.12	所提方法	KITTI 07	0.04	0.51	102.17	5.83
LeGO-LOAM	KITTI 02	12.91	60.52	102.60	6.34	所提方法	KITTI 07	0.04	0.51	102.17	5.83
LeGO-LOAM	KITTI 02	12.91	60.52	102.60	6.34	所提方法	KITTI 09	0.07	0.70	101.77	5.50
LeGO-LOAM	KITTI 05	1.31	3.89	105.04	5.36	所提方法	KITTI 09	0.07	0.70	101.77	5.50
LeGO-LOAM	KITTI 05	1.31	3.89	105.04	5.36	所提方法	本地数据	0.04	0.38	46.97	5.54
LeGO-LOAM	KITTI 06	0.91	3.05	103.79	5.12	所提方法	本地数据	0.04	0.38	46.97	5.54

[1]	赖国良, 胡钊政, 周哲, 万金杰, 任靖渊. 基于语义似然与高精度地图匹配的智能车辆同时定位与检测[J]. 上海交通大学学报, 2024, 58(10): 1618-1628.
[2]	李翠明, 王华, 徐龙儿, 王龙. 基于改进DeepLabv3+的光伏电站道路识别方法[J]. 上海交通大学学报, 2024, 58(5): 776-782.
[3]	庄浩宇, 颜国正, 费倩, 汪炜, 赵凯. 肠道机器人无线供能的混合式三维发射线圈特性研究[J]. 上海交通大学学报, 2023, 57(5): 545-551.
[4]	林淳, 贺越生, 方兴其, 王春香. 一种面向泊车场景智能车辆轨迹规划方法[J]. 上海交通大学学报, 2023, 57(3): 345-353.
[5]	黄宇昊, 韩超, 赵明辉, 杜乾坤, 王石刚. 考虑安全飞行通道约束的无人机飞行轨迹多目标优化策略[J]. 上海交通大学学报, 2022, 56(8): 1024-1033.
[6]	齐东润, 陈刚. 无人驾驶机器人多目标模糊操纵策略[J]. 上海交通大学学报, 2021, 55(10): 1310-1319.
[7]	薛蓉蓉, 王志武, 颜国正, 庄浩宇. 肠道机器人获取的肠道图像降噪处理方法[J]. 上海交通大学学报, 2021, 55(10): 1303-1309.
[8]	胡益恺, 庄瀚洋, 王春香, 杨明. 基于主从博弈的智能车汇流场景决策方法[J]. 上海交通大学学报, 2021, 55(8): 1027-1034.
[9]	胡益恺, 王春香, 杨明. 智能车辆决策方法研究综述[J]. 上海交通大学学报, 2021, 55(8): 1035-1048.
[10]	刘忠, 曹其新, 朱笑笑, 王雯珊. 基于VRML节点树的双臂移动机器人碰撞检测及优化[J]. 上海交通大学学报（自然版）, 2011, 45(07): 985-989.
[11]	陈平a, 王春香a, 杨明b, 王冰b. 基于随动预警区域的主动安全避碰系统设计[J]. 上海交通大学学报（自然版）, 2011, 45(07): 995-999.
[12]	梁昆，杨明，王春香，王冰. 基于低精度GPS的智能车视觉导航方法 [J]. 上海交通大学学报（自然版）, 2011, 45(02): 168-0172.
[13]	王景川,陈卫东,张栩. 一种鲁棒的室外移动机器人路标定位系统 [J]. 上海交通大学学报（自然版）, 2010, 44(09): 1181-1186.
[14]	张宇腾，杨明，王春香. 基于主动视觉的智能车道路跟踪方法 [J]. 上海交通大学学报（自然版）, 2010, 44(08): 1042-1045.