Loop Closure Detection Method of Laser SLAM Based on Global Feature Descriptor

doi:10.16183/j.cnki.jsjtu.2021.202

Abstract

Abstract:

To solve the problem that localization error of the underground inspection system continues to accumulate over time, a loop closure detection algorithm based on point cloud global feature descriptor is proposed, which is suitable for laser simultaneous localization and mapping (SLAM). The feature vector of each point in point cloud is calculated by curvature, then the global feature descriptor of point cloud is constructed based on the angle distribution and scale distribution relationship between the feature vector and center point coordinate system. In addition, the pose transformation of two similar frames is calculated by feature point registration to improve computing efficiency. The proposed algorithm is verified by simulation experiments and open-source data experiments. The experimental results show that the proposed algorithm has a significant improvement in localization accuracy and real-time performance, which can effectively solve the problems of increased cumulative error and poor global consistency of the localization algorithm during long-term inspections.

Key words: underground inspection, laser localization, loop closure detection, global feature descriptor

CLC Number:

TP242

HAN Chao, CHEN Min, HUANG Yuhao, ZHAO Minghui, DU Qiankun, LIANG Qinhua. Loop Closure Detection Method of Laser SLAM Based on Global Feature Descriptor[J]. Journal of Shanghai Jiao Tong University, 2022, 56(10): 1379-1387.

Figures/Tables 12

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Fig.7

Fig.8

Tab.1

Fig.9

Tab.2

Fig.10

References 19

[1]	裴福俊, 李昊洋, 武玫. 基于分布式无味边缘粒子滤波的同步定位与地图构建[J]. 上海交通大学学报, 2014, 48(7): 987-992.
	PEI Fujun, LI Haoyang, WU Mei. Distributed unscented marginalized particle filter for simultaneous localization and mapping[J]. Journal of Shanghai Jiao Tong University, 2014, 48(7): 987-992.
[2]	AZHARI F, KIELY S, SENNERSTEN C, et al. A comparison of sensors for underground void mapping by unmanned aerial vehicles[C]//Proceedings of the First International Conference on Underground Mining Technology. Perth, Australia: ACG, 2017: 419-430.
[3]	杨晶东, 彭坤, 顾浩楠, 等. 基于直线特征的机器人自主定位方法[J]. 上海交通大学学报, 2018, 52(9): 1120-1124.
	YANG Jingdong, PENG Kun, GU Haonan, et al. Autonomous localisation method based on linear feature for robots[J]. Journal of Shanghai Jiao Tong University, 2018, 52(9): 1120-1124.
[4]	REN Z L, WANG L G, BI L. Robust GICP-based 3D LiDAR SLAM for underground mining environment[J]. Sensors, 2019, 19(13): 2915. doi: 10.3390/s19132915 URL
[5]	LI M G, ZHU H, YOU S Z, et al. Efficient laser-based 3D SLAM for coal mine rescue robots[J]. IEEE Access, 2019, 7: 14124-14138. doi: 10.1109/ACCESS.2018.2889304
[6]	MUR-ARTAL R, TARDÓS J D. ORB-SLAM2: An open-source SLAM system for monocular, stereo, and RGB-D cameras[J]. IEEE Transactions on Robotics, 2017, 33(5): 1255-1262. doi: 10.1109/TRO.2017.2705103 URL
[7]	QIN T, LI P L, SHEN S J. VINS-mono: A robust and versatile monocular visual-inertial state estimator[J]. IEEE Transactions on Robotics, 2018, 34(4): 1004-1020. doi: 10.1109/TRO.2018.2853729 URL
[8]	GALVEZ-LÓPEZ D, TARDOS J D. Bags of binary words for fast place recognition in image sequences[J]. IEEE Transactions on Robotics, 2012, 28(5): 1188-1197. doi: 10.1109/TRO.2012.2197158 URL
[9]	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 and Systems. Madrid, Spain:IEEE, 2018: 4758-4765.
[10]	SHAN T X, ENGLOT B, MEYERS D, et al. LIO-SAM: Tightly-coupled lidar inertial odometry via smoothing and mapping[C]//2020 IEEE/RSJ International Conference on Intelligent Robots and Systems. Las Vegas, NV, USA: IEEE, 2021: 5135-5142.
[11]	柴梦娜, 刘元盛, 任丽军. 基于激光点云NDT特征的两步回环检测[J]. 激光与红外, 2020, 50(1): 17-24.
	CHAI Mengna, LIU Yuansheng, REN Lijun. Two-step loop closure detection based on laser point cloud NDT features[J]. Laser & Infrared, 2020, 50(1): 17-24.
[12]	李旭. 基于多线激光雷达建图的里程计优化及回环检测[D]. 哈尔滨: 哈尔滨工业大学, 2019.
	LI Xu. Odometry optimization and loop detection based on multi-beam LiDAR mapping[D]. Harbin:Harbin Institute of Technology, 2019.
[13]	ZHANG J, SINGH S. LOAM: Lidar odometry and mapping in real-time[EB/OL]. (2014-07-16) [2021-05-18]. http://www.roboticsproceedings.org/rss10/p07.pdf.
[14]	LIU M, POMERLEAU F, COLAS F, et al. Normal estimation for pointcloud using GPU based sparse tensor voting[C]//2012 IEEE International Conference on Robotics and Biomimetics. Guangzhou, China: IEEE, 2012: 91-96.
[15]	STIGLER S M. Francis galton’s account of the invention of correlation[J]. Statistical Science, 1989, 4(2): 73-79.
[16]	张小利, 李雄飞, 李军. 融合图像质量评价指标的相关性分析及性能评估[J]. 自动化学报, 2014, 40(2): 306-315.
	ZHANG Xiaoli, LI Xiongfei, LI Jun. Validation and correlation analysis of metrics for evaluating performance of image fusion[J]. Acta Automatica Sinica, 2014, 40(2): 306-315.
[17]	WEN W S, KAN Y C, HSU L T. Performance comparison of GNSS/INS integrations based on EKF and factor graph optimization[C]//Proceedings of the 32nd International Technical Meeting of the Satellite Division of The Institute of Navigation. Miami, Florida, USA: Institute of Navigation, 2019: 3019-3032.
[18]	RATZ S, DYMCZYK M, SIEGWART R, et al. OneShot global localization: Instant LiDAR-visual pose estimation[C]//2020 IEEE International Conference on Robotics and Automation. Paris, France: IEEE, 2020: 5415-5421.
[19]	DANG T, KHATTAK S, MASCARICH F, et al. Explore locally, plan globally: A path planning framework for autonomous robotic exploration in subterranean environments[C]//2019 19th International Conference on Advanced Robotics. Belo Horizonte, Brazil: IEEE, 2019: 9-16.

序号	实验算法	最大误差/m	终点误差/m	均方根误差/m	巡检距离/m
1	LOAM	3.48863	3.45962	1.52933	618.007
	LOAM+ICP	2.50563	0.14402	1.03654	618.007
	本文算法	1.46447	0.06376	0.58827	618.007
2	LOAM	1.69615	1.53225	0.83298	638.446
	LOAM+ICP	0.54603	0.02760	0.32963	638.446
	本文算法	0.40609	0.02806	0.17783	638.446
3	LOAM	1.16494	1.15254	0.71497	264.852
	LOAM+ICP	0.82921	0.08813	0.43781	264.852
	本文算法	0.47062	0.09923	0.21525	264.852
4	LOAM	1.92765	0.16456	1.06969	197.782
	LOAM+ICP	0.45537	0.04628	0.22482	197.782
	本文算法	0.39201	0.05806	0.15977	197.782

算法	相似度检测时间/ms	位姿计算时间/ms	回环时间/ms
LOAM+ICP	12.6676	122.3668	147.0491
本文算法	16.2595	67.4053	88.7274