一种高动态场景下视觉激光融合SLAM系统

doi:10.19562/j.chinasae.qcgc.2025.09.007

摘要/Abstract

摘要：

精确的定位与建图是无人驾驶系统的关键，依赖于单一传感器的同时定位与建图（SLAM）系统很难在不同的环境中稳定运行，特别是高动态场景，容易受动态障碍物影响导致精度下降甚至失效。为此，本文提出了一种激光视觉融合SLAM框架解决高动态场景下高精度建图与定位问题。首先，设计了一种稀疏激光点云与稠密图像融合的里程计，充分利用激光的高精度测距特性和图像信息丰富的特点，提高了里程计精度。针对高动态场景，基于实时图像语义分割网络BiSeNetV2与基于帧间与连续多帧的运动特征检测技术，精确高效地检测图像激光融合后的3D特征点中的动态点，将其从地图中去除，消除动态障碍物的影响。在无人驾驶数据集nuScenes上进行测试，测试结果表明所提出系统大幅提高了动态场景下定位建图的精度和鲁棒性。

关键词: 同时定位与建图, 图像, 激光雷达, 多传感器融合, 高动态场景

Abstract:

Accurate localization and mapping are critical for autonomous driving systems. However， single-sensor Simultaneous Localization and Mapping （SLAM） systems often struggle to operate reliably across different environment， particularly in highly dynamic scenes where dynamic obstacles can degrade accuracy or even cause system failure. Therefore， in this paper a LiDAR-Visual fusion SLAM framework tailored for high-precision mapping and localization problems in dynamic environment is proposed. Firstly， an odometry method that fuses sparse LiDAR point clouds with dense image data is designed， leveraging the high-precision ranging capabilities of LiDAR and the rich information provided by images to enhance odometry accuracy. To address challenges in highly dynamic scenes， based on a real-time image semantic segmentation network， BiSeNetV2， combined with motion feature detection techniques based on inter-frame and multi-frame sequences， the efficient and accurate identification of dynamic points among the 3D feature points obtained from the LiDAR-Visual fusion is realized， which are removed from the map to mitigate the influence of dynamic obstacles. Tests are carried out on the nuScenes autonomous driving dataset， and the results show significant improvement of the proposed system in accuracy and robustness of localization and mapping in dynamic environment.

Key words: SLAM, image, LiDAR, multi-sensor fusion, high-dynamic environment

周云水,高澄宇,黄圣杰,张润邦,陈新,边有钢,秦洪懋. 一种高动态场景下视觉激光融合SLAM系统[J]. 汽车工程, 2025, 47(9): 1712-1720.

Yunshui Zhou,Chengyu Gao,Shengjie Huang,Runbang Zhang,Xin Chen,Yougang Bian,Hongmao Qin. LiDAR-Visual Fusion SLAM System for High-Dynamic Environment[J]. Automotive Engineering, 2025, 47(9): 1712-1720.

图/表 9

图1

图2

图3

图4

图5

表 1

图6

表2

表3

参考文献 26

[1]	CAMPOS C， ELVIRA R， RODRÍGUEZ J J G， et al. ORB-SLAM3： an accurate open-source library for visual， visual-inertial， and multimap SLAM［J］. IEEE Transactions on Robotics， 2021， 37（6）： 1874-1890.
[2]	ENGEL J， KOLTUN V， CREMERS D. Direct sparse odometry［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2017， 40（3）： 611-625.
[3]	FORSTER C， PIZZOLI M， SCARAMUZZA D. SVO： fast semi-direct monocular visual odometry［C］. 2014 IEEE International Conference on Robotics and Automation （ICRA）. IEEE， 2014： 15-22.
[4]	杨蒙蒙，江昆，温拓朴，等. 自动驾驶高精度地图众源更新技术现状与挑战［J］. 中国公路学报， 2023， 36（5）： 244-259.
	YANG M， JIANG K， WEN T， et al. Review on status and challenges of crowdsourced updating of highly automated driving maps ［J］. China Journal of Highway and Transport， 2023， 36（5）： 244-259.
[5]	YAN M， WANG J， LI J， et al. Loose coupling visual-lidar odometry by combining VISO2 and LOAM［C］. 2017 36th Chinese Control Conference （CCC）. IEEE， 2017： 6841-6846.
[6]	XU Y， OU Y， XU T. SLAM of robot based on the fusion of vision and LIDAR［C］. 2018 IEEE International Conference on Cyborg and Bionic Systems （CBS）. IEEE， 2018： 121-126.
[7]	ZHANG J， SINGH S. Visual-lidar odometry and mapping： low-drift， robust， and fast［C］. 2015 IEEE International Conference on Robotics and Automation （ICRA）. IEEE， 2015： 2174-2181.
[8]	ZUO X， GENEVA P， LEE W， et al. LIC-Fusion： LIDAR-inertial-camera odometry［C］. 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems （IROS）. IEEE， 2019： 5848-5854.
[9]	SUN K， MOHTA K， PFROMMER B， et al. Robust stereo visual inertial odometry for fast autonomous flight［J］. IEEE Robotics and Automation Letters， 2018， 3（2）： 965-972.
[10]	ZUO X， YANG Y， GENEVA P， et al. LIC-Fusion 2.0： LIDAR-inertial-camera odometry with sliding-window plane-feature tracking［C］. 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems （IROS）. IEEE， 2020： 5112-5119.
[11]	WISTH D， CAMURRI M， DAS S， et al. Unified multi-modal landmark tracking for tightly coupled lidar-visual-inertial odometry［J］. IEEE Robotics and Automation Letters， 2021， 6（2）： 1004-1011.
[12]	LIN J， ZHANG F. R3LIVE： a robust， real-time， RGB-colored， LiDAR-Inertial-Visual tightly-coupled state estimation and mapping package［C］. 2022 International Conference on Robotics and Automation （ICRA）. IEEE， 2022： 10672-10678.
[13]	LIN J， ZHANG F. R3LIVE++： a robust， real-time， radiance reconstruction package with a tightly-coupled LiDAR-Inertial-Visual state estimator［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2024.
[14]	ZOU D， TAN P. Coslam： collaborative visual SLAM in dynamic environments［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2012， 35（2）： 354-366.
[15]	KUNDU A， KRISHNA K M， SIVASWAMY J. Moving object detection by multi-view geometric techniques from a single camera mounted robot［C］. 2009 IEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE， 2009： 4306-4312.
[16]	YANG X， YUAN Z， ZHU D， et al. Robust and efficient RGB-D SLAM in dynamic environments［J］. IEEE Transactions on Multimedia， 2020， 23： 4208-4219.
[17]	KLAPPSTEIN J， VAUDREY T， RABE C， et al. Moving object segmentation using optical flow and depth information［C］. Advances in Image and Video Technology： Third Pacific Rim Symposium， PSIVT 2009， Tokyo， Japan， January 13-16， 2009. Proceedings 3. Springer Berlin Heidelberg， 2009： 611-623.
[18]	YU C， LIU Z， LIU X J， et al. DS-SLAM： a semantic visual SLAM towards dynamic environments［C］. 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems （IROS）. IEEE， 2018： 1168-1174.
[19]	陈吉清，车宇翔，田小强，等.动态场景下基于3D多目标追踪的实时视觉SLAM方法研究［J］.汽车工程，2024，46（5）：776-783.
	CHEN J Q， CHE Y X， TIAN X Q， et al. Research on real-time visual SLAM method based on 3D multi-object tracking in dynamic scenes［J］. Automotive Engineering， 2024， 46（5）： 776-783.
[20]	刘钰嵩，何丽，袁亮，等.动态场景下基于光流的语义RGBD-SLAM算法［J］.仪器仪表学报，2022，43（12）：139-148.
	LIU Y S， HE L， YUAN L， et al. Semantic RGBD-SLAM in dynamic scene based on optical flow［J］. Chinese Journal of Scientific Instrument， 2022， 43（12）：139-148.
[21]	YU C， GAO C， WANG J， et al. BiSeNet V2： bilateral network with guided aggregation for real-time semantic segmentation［J］. International Journal of Computer Vision， 2021， 129： 3051-3068.
[22]	LANCASTER P. Algebraic riccati equations［M］. Oxford Science Publications/The Clarendon Press， Oxford University Press， 1995.
[23]	CORDTS M， OMRAN M， RAMOS S， et al. The cityscapes dataset for semantic urban scene understanding［C］. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition， 2016： 3213-3223.
[24]	ZHOU B， ZHAO H， PUIG X， et al. Semantic understanding of scenes through the ADE20K dataset［J］. International Journal of Computer Vision， 2019， 127： 302-321.
[25]	CAESAR H， BANKITI V， LANG A H， et al. nuScenes： a multimodal dataset for autonomous driving［C］. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition， 2020： 11621-11631.
[26]	秦晓辉，周洪，廖毅霏，等. 动态环境下基于时序滑动窗口的鲁棒激光SLAM系统［J］. 湖南大学学报（自然科学版）， 2023， 12： 49-58.
	QIN X H， ZHOU H， LIAO Y F， et al. Robust laser SLAM system based on temporal sliding window in dynamic scenes［J］. Journal of Hunan University （Natural Sciences）， 2023， 12： 49-58.

数据集	480×640	1280×720	精度下降率
Cityscape	7.38 ms	12.59 ms	0.67 %
ADE20K	7.74 ms	15.03 ms	2.89 %

序列	本文里程计			激光语义过滤方法			本文方法
序列	RMSE/m	Mean/m	STD	RMSE/m	Mean/m	STD	RMSE/m	Mean/m	STD
103	1.677	1.373	0.038	0.851	0.785	0.227	0.537	0.473	0.093
553	1.191	0.894	0.029	0.015	0.014	0.002	0.073	0.016	0.079
757	1.354	1.371	0.138	0.316	0.308	0.069	0.711	0.201	1.337
1 077	2.428	2.031	1.751	3.482	2.499	2.425	1.871	0.968	1.114
1 094	1.876	1.621	0.025	1.476	0.966	1.117	1.124	1.109	0.111
1 100	0.925	0.725	0.033	0.081	0.067	0.045	0.073	0.063	0.017

序列	RMSE/m
序列	ORBSlam3	VLOAM	本文里程计	里程计+语义
103	2.815	0.521	1.677	0.537
553	1.261	0.217	1.191	0.073
757	2.775	2.376	1.354	0.711
1 077	2.643	2.438	2.428	1.871
1 094	1.928	1.531	1.876	1.124
1 100	1.072	0.331	0.925	0.073