轻量的增强型激光雷达-惯性-视觉里程计系统

doi:10.13474/j.cnki.11-2246.2025.0913

摘要/Abstract

摘要： 激光-惯性-视觉里程计(LIVO)在移动机器人和自动驾驶等领域展现出广泛的应用潜力。本文基于FAST-LIVO提出了一种轻量的增强型激光雷达-惯性-视觉里程计系统——LITE-LIVO。该系统通过集成激光雷达、惯性测量单元(IMU)和视觉传感器,实现高效且实时的姿态估计与高精度地图构建;为提高系统在动态光照条件下的稳健性,引入一种基于深度学习的特征点提取方法和稀疏光流跟踪方法,并通过构建视觉观测残差,在卡尔曼滤波中融合视觉与激光雷达信息;此外,设计了紧耦合的视觉-惯性里程计(VIO)子系统,从激光雷达点云中筛选高质量视觉特征,同时更有效地管理视觉地图。试验结果表明,LITE-LIVO在多个公开数据集和实际场景中均表现出色,尤其在处理复杂环境和退化场景时展现了显著的优势。本文为激光-惯性-视觉里程计的发展提供了新的思路和方法,提高了多源数据融合的定位精度,增加了移动机器人的应用场景。

关键词: 激光-惯性-视觉里程计, 深度学习, 光流跟踪, 卡尔曼滤波

Abstract: Lightweight intelligent tracking for enhanced LiDAR-IMU-visual odometry(LIVO)has broad applications in mobile robotics and autonomous driving.This paper proposes LITE-LIVO,a lightweight and enhanced LIVO system built upon FAST-LIVO,integrating LiDAR,IMU and vision sensors for real-time pose estimation and high-precision mapping.To enhance robustness under dynamic lighting,the system employs a deep learning-based feature extraction and sparse optical flow tracking,fusing visual and LiDAR data via Kalman filtering with visual residuals.A tightly coupled visual-IMU odometry (VIO)subsystem filters high-quality visual features from LiDAR point clouds and optimizes visual map management.Experimental results on public datasets and real-world scenarios demonstrate superior performance,especially in complex and degraded environments.This study advances multi-source data fusion techniques,improving localization accuracy and expanding application domains for mobile robots.

Key words: LiDAR-inertial-visual odometry, deep learning, optical flow tracking, Kalman filter

中图分类号:

P237

杨颜光, 钱建国, 于斌, 郭洁, 焦扬. 轻量的增强型激光雷达-惯性-视觉里程计系统[J]. 测绘通报, 2025, 0(9): 78-83,104.

YANG Yanguang, QIAN Jianguo, YU Bin, GUO Jie, JIAO Yang. A lightweight enhanced LiDAR-inertial-visual odometry system[J]. Bulletin of Surveying and Mapping, 2025, 0(9): 78-83,104.

参考文献

[1]DURRANT-WHYTE H, BAILEY T.Simultaneous localization and mapping: part I[J].IEEE Robotics ＆ Automation Magazine, 2006, 13(2):99-110.
[2]KHAN M U, ALI ZAIDI S A, ISHTIAQ A, et al.A comparative survey of LiDAR-SLAM and LiDAR based sensor technologies[C]//Proceedings of 2021 Mohammad Ali Jinnah University International Conference on Computing (MAJICC).Karachi:IEEE, 2021:1-8.
[3]HUANG, L.Review on LiDAR-based SLAM techniques[C]//Proceedings of 2021 International Conference on Signal Processing and Machine Learning (CONF-SPML).[S.l]:IEEE, 2021:163-168.
[4]王晨捷, 罗斌, 李成源, 等.无人机视觉SLAM协同建图与导航[J].测绘学报, 2020, 49(6):767-776.
[5]TAKETOMI T, UCHIYAMA H, IKEDA S.Visual SLAM algorithms:a survey from 2010 to 2016[J].IPSJ Transactions on Computer Vision and Applications, 2017, 9(1):16.
[6]ZHANG J, SINGH S.LOAM:LiDAR odometry and mapping in real-time[J].Robotics:Science and Systems Foundation, 2014, 2(9):1-9.
[7]SHAN T X, ENGLOT B.LeGO-LOAM:lightweight and ground-optimized lidar odometry and mapping on variable terrain[C]//Proceedings of 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).Madrid:IEEE, 2018:4758-4765.
[8]SHAN T X, ENGLOT B, MEYERS D, et al.LIO-SAM:tightly-coupled LiDAR inertial odometry via smoothing and mapping[C]//Proceedings of 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).Las Vegas, NV:IEEE, 2020:5135-5142.
[9]XU W, ZHANG F.FAST-LIO:a fast, robust LiDAR-inertial odometry package by tightly-coupled iterated Kalman filter[EB/OL].[2024-12-20].https://arxiv.org/abs/2010.08196v3.
[10] XU W, CAI Y X, HE D J, et al.FAST-LIO2:fast direct LiDAR-inertial odometry[J].IEEE Transactions on Robotics, 2022, 38(4):2053-2073.
[11] 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.
[12] MUR-ARTAL R, TARDOS 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.
[13] XU K, HAO Y F, YUAN S H, et al.AirSLAM:an efficient and illumination-robust point-line visual SLAM system[EB/OL].[2024-12-20].https://arxiv.org/abs/2408.03520v4.
[14] LI R H, WANG S, LONG Z Q, et al.UnDeepVO:monocular visual odometry through unsupervised deep learning[EB/OL].[2024-12-20].https://arxiv.org/abs/1709.06841v2.
[15] ZHENG C, ZHU Q, XU W, et al.Fast-livo:Fast and tightly-coupled sparse-direct lidar-inertial-visual odometry[C]//Proceedings of 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).[S.l]:IEEE, 2022:4003-4009.
[16] BAILEY T, NIETO J, GUIVANT J, et al.Consistency of the EKF-SLAM algorithm[C]//Proceedings of 2006 IEEE/RSJ International Conference on Intelligent Robots and Systems.Beijing:IEEE, 2006:3562-3568.
[17] 于宁波, 王石荣.利用双RGB-D传感器融合增强对未知环境的自主探索和地图构建[J].工程(英文), 2019, 5(1):355-373.
[18] XUE J, WANG D, DU S, CUI D, et al.Visual-dominant multi-sensor fusion method for autonomous localization and obstacle perception in autonomous vehicles[J].Frontiers of Information Technology ＆ Electronic Engineering, 2017, 1:122-139.
[19] 文成林, 郭超, 高敬礼.多传感器多尺度图像信息融合算法[J].电子学报, 2008, 36(5):840-847.
[20] 高强, 陆科帆, 吉月辉, 等.多传感器融合SLAM研究综述[J].现代雷达, 2024, 46(8):29-39.
[21] RONNEBERGER O, FISCHER P, BROX T.U-Net:convolutional networks for biomedical image segmentation[M]//Medical Image Computing and Computer-Assisted Intervention—MICCAI 2015.Cham:Springer International Publishing, 2015:234-241.
[22] LIN Y C, WANG S, JIANG Y L, et al.Breaking of brightness consistency in optical flow with a lightweight CNN network[J].IEEE Robotics and Automation Letters, 2024, 9(8):6840-6847.
[23] NGUYEN T M, YUAN S H, CAO M Q, et al.NTU VIRAL:a visual-inertial-ranging-LiDAR dataset, from an aerial vehicle viewpoint[J].The International Journal of Robotics Research, 2022, 41(3):270-280.
[24] LIN J R, ZHANG F.R³ LIVE:a robust, real-time, RGB-colored, LiDAR-inertial-visual tightly-coupled state estimation and mapping package[C]//Proceedings of 2022 International Conference on Robotics and Automation (ICRA).Philadelphia, PA:IEEE, 2022:10672-10678.