多相机组合式全景视觉惯性SLAM系统

doi:10.12305/j.issn.1001-506X.2026.08.26

摘要/Abstract

摘要：

针对室外复杂场景下，视觉定位算法精度和鲁棒性下降、低成本惯性测量单元(inertial measurement unit, IMU)误差快速累积导致姿态漂移等问题，提出一种全景视觉惯性同时定位与成图(panoramic visual-inertial simultaneous localization and mapping, PVI-SLAM)算法。在系统架构层面，将惯性状态深度嵌入到追踪、建图及闭环等模块，构建视觉惯性融合框架。在全景初始化的基础上，结合参数解耦策略与解析求解方法计算参数初值并采用最大后验概率估计完成惯导初始化，实现视觉与惯性状态空间的精确对齐。同时，将全景视觉重投影误差与IMU预积分约束统一建模为非线性最小二乘问题，采用图优化方法完成状态估计。实验验证表明，在复杂室外场景下，PVI-SLAM算法相比现有主流SLAM方案，在状态对齐性能、定位精度和追踪鲁棒性等方面均展现出显著优势。

关键词: 同时定位与成图, 多相机, 全景视觉惯性对齐, 状态估计

Abstract:

To address the issues of decreased accuracy and robustness of visual localization algorithms in complex outdoor scenarios, as well as attitude drift caused by the rapid error accumulation of low-cost inertial measurement unit (IMU), a panoramic visual-inertial simultaneous localization and mapping (PVI-SLAM) algorithm is proposed. At the system architecture level, inertial states are deeply integrated into modules such as tracking, mapping, and loop closing to construct a visual-inertial fusion framework. Building on panoramic initialization, a parameter decoupling strategy and an analytical solution method are combined to calculate initial parameter values. It further employs maximum a posteriori estimation to complete inertial navigation initialization, thereby achieving accurate alignment between the visual and inertial state spaces. Meanwhile, the panoramic visual reprojection error and IMU pre-integration constraints are uniformly modeled as a nonlinear least squares problem, and the graph optimization method is used to complete state estimation. Experimental verification demonstrates that in complex outdoor scenarios, compared with existing mainstream SLAM solutions, the PVI-SLAM algorithm exhibits significant advantages in terms of state alignment performance, localization accuracy, and tracking robustness.

Key words: simultaneous localization and mapping (SLAM), multi-camera, panoramic visual-inertial alignment, state estimation

中图分类号:

TP 242

卓德胜, 朱锋, 张小红, 程军龙, 胡捷. 多相机组合式全景视觉惯性SLAM系统[J]. 系统工程与电子技术, 2026, 48(8): 2809-2820.

Desheng ZHUO, Feng ZHU, Xiaohong ZHANG, Junlong CHENG, Jie HU. Multi-camera combined panoramic visual-inertial SLAM system[J]. Systems Engineering and Electronics, 2026, 48(8): 2809-2820.

图/表 15

图1

图2

图3

图4

图5

图6

表1

图7

表2

表3

表4

图8

表5

图9

表6

参考文献 31

1	ABASPUR K I, FITZGERALD L, DOOLY G, et al. A survey of state-of-the-art on visual SLAM[J]. Expert Systems with Applications, 2022, 205, 117734. doi: 10.1016/j.eswa.2022.117734
2	LIU Y P, JIA T, GUO S Y, et al. A robust and accurate stereo SLAM based on learned feature extraction and matching[J]. IEEE Trans. on Instrumentation and Measurement, 2025, 74, 8514012.
3	AL-TAWIL B, HEMPEL T, ABDELRAHMAN A, et al. A review of visual SLAM for robotics: evolution, properties, and future applications[J]. Frontiers in Robotics and AI, 2024, 11, 1347985. doi: 10.3389/frobt.2024.1347985
4	MUR-ARTAL R, MONTIEL J M M, TARDOS J D. ORB-SLAM: a versatile and accurate monocular SLAM system[J]. IEEE Trans. on Robotics, 2015, 31 (5): 1147- 1163. doi: 10.1109/TRO.2015.2463671
5	MUR-ARTAL R, TARDOS J D. ORB-SLAM2: an open-source SLAM system for monocular, stereo and RGB-D cameras[J]. IEEE Trans. on Robotics, 2017, 33 (5): 1255- 1262. doi: 10.1109/TRO.2017.2705103
6	ZHENG J H, ZHU Z H, BIERI V, et al. WildGS-SLAM: monocular Gaussian splatting SLAM in dynamic environments[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2025: 11461−11471.
7	MO J, ISLAM M J, SATTAR J. Fast direct stereo visual SLAM[J]. IEEE Robotics and Automation Letters, 2022, 7 (2): 778- 785. doi: 10.1109/LRA.2021.3133860
8	XU K, HAO Y F, YUAN S H, et al. AirVO: an illumination-robust point-line visual odometry[C]//Proc. of the IEEE/RSJ International Conference on Intelligent Robots and Systems, 2023: 3429−3436.
9	CAO Z Z, ZHAO C Y, ZHANG Q Y, et al. RGBDS-SLAM: a RGB-D semantic dense SLAM based on 3D multi level pyramid Gaussian splatting[J]. IEEE Robotics and Automation Letters, 2025, 10 (5): 4778- 4785. doi: 10.1109/LRA.2025.3553049
10	TITKOV R, ZUBKOV E, YUDIN D, et al. LEG-SLAM: real-time language-enhanced Gaussian splatting for SLAM[EB/OL]. [2025-09-10]. http://arxiv.org/pdf/2506.03073.
11	董荣, 厉茂海, 林睿, 等. 多相机与IMU融合的室外机器人定位方法研究[J]. 计算机工程与应用, 2022, 58 (3): 289- 296. doi: 10.3778/j.issn.1002-8331.2008-0257
	DONG R, LI M H, LIN R, et al. Research on outdoor robot positioning method fusing multi-camera and IMU[J]. Computer Engineering and Applications, 2022, 58 (3): 289- 296. doi: 10.3778/j.issn.1002-8331.2008-0257
12	季顺平, 秦梓杰. 多镜头组合式相机的全景SLAM[J]. 测绘学报, 2019, 48 (10): 1254- 1265.
	JI S P, QIN Z J. Panoramic SLAM with multi-lens combined cameras[J]. Acta Geodaetica et Cartographica Sinica, 2019, 48 (10): 1254- 1265.
13	FORSTER C, ZHANG Z C, GASSNER M, et al. SVO: semidirect visual odometry for monocular and multicamera systems[J]. IEEE Trans. on Robotics, 2017, 33 (2): 249- 265. doi: 10.1109/TRO.2016.2623335
14	JI S P, QIN Z J, SHAN J, et al. Panoramic SLAM from a multiple fisheye camera rig[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2020, 159, 169- 183. doi: 10.1016/j.isprsjprs.2019.11.014
15	KAVETI P, VAIDYANATHAN S N, CHELVAN A T, et al. Design and evaluation of a generic visual SLAM framework for multi camera systems[J]. IEEE Robotics and Automation Letters, 2023, 8 (11): 7368- 7375. doi: 10.1109/LRA.2023.3316609
16	PLESS R. Using many cameras as one[C]//Proc. of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003.
17	SONG H, LIU C, DAI H F. BundledSLAM: an accurate visual SLAM system using multiple cameras[C]//Proc. of the IEEE 7th Advanced Information Technology, Electronic and Automation Control Conference, 2024: 106−111.
18	章国锋, 黄赣, 谢卫健, 等. 基于单目视觉惯性的同步定位与地图构建方法综述[J]. 中国图象图形学报, 2024, 29 (10): 2839- 2858. doi: 10.11834/jig.230863
	ZHANG G F, HUANG G, XIE W J, et al. A review of monocular visual-inertial simultaneous localization and mapping methods[J]. Journal of Image and Graphics, 2024, 29 (10): 2839- 2858. doi: 10.11834/jig.230863
19	HERRERA-GRANDA E P, TORRES-CANTERO J C, PELUFFO-ORDÓÑEZ D H. Monocular visual SLAM, visual odometry, and structure from motion methods applied to 3D reconstruction: a comprehensive survey[J]. Heliyon, 2024, 10 (18): e37356. doi: 10.1016/j.heliyon.2024.e37356
20	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. doi: 10.1109/LRA.2018.2793349
21	QIN T, LI P L, SHEN S J. VINS-Mono: a robust and versatile monocular visual-inertial state estimator[J]. IEEE Trans. on Robotics, 2018, 34 (4): 1004- 1020. doi: 10.1109/TRO.2018.2853729
22	GENEVA P, ECKENHOFF K, LEE W, et al. OpenVINS: a research platform for visual-inertial estimation[C]//Proc. of the IEEE International Conference on Robotics and Automation, 2020: 4666−4672.
23	CAMPOS C, ELVIRA R, RODRÍGUEZ J J G, et al. ORB-SLAM3: an accurate open-source library for visual, visual-inertial and multi-map SLAM[J]. IEEE Trans. on Robotics, 2021, 37 (6): 1874- 1890. doi: 10.1109/TRO.2021.3075644
24	刘悦, 李化义, 张世杰, 等. 面向视觉惯导的导航系统初始化技术综述[J]. 计算机工程与应用, 2025, 61 (2): 1- 18. doi: 10.3778/j.issn.1002-8331.2406-0038
	LIU Y, LI H Y, ZHANG S J, et al. A survey of techniques for visual-inertial navigation system initialization[J]. Computer Engineering and Applications, 2025, 61 (2): 1- 18. doi: 10.3778/j.issn.1002-8331.2406-0038
25	JI S P, SHI Y, SHI Z C, et al. Comparison of two panoramic sensor models for precise 3D measurements[J]. Photogrammetric Engineering & Remote Sensing, 2014, 80 (3): 229- 238. doi: 10.14358/PERS.80.3.229
26	KUMMERLE R, GRISETTI G, STRASDAT H, et al. G²o: a general framework for graph optimization[C]//Proc. of the IEEE International Conference on Robotics and Automation, 2011: 3607−3613.
27	MUR-ARTAL R, TARDOS J D. Visual-inertial monocular SLAM with map reuse[J]. IEEE Robotics and Automation Letters, 2017, 2 (2): 796- 803. doi: 10.1109/LRA.2017.2653359
28	FORSTER C, CARLONE L, DELLAERT F, et al. On-manifold preintegration for real-time visual--inertial odometry[J]. IEEE Trans. on Robotics, 2017, 33(1): 1−21.
29	ZUNIGA-NOEL D, MORENO F A, GONZALEZ-JIMENEZ J. An analytical solution to the IMU initialization problem for visual-inertial systems[J]. IEEE Robotics and Automation Letters, 2021, 6 (3): 6116- 6122. doi: 10.1109/LRA.2021.3091407
30	ZDRAVKOVSKA N. CRC standard mathematical tables and formulas[M]. New York: Chapman and Hall/CRC, 2018.
31	UMEYAMA S. Least-squares estimation of transformation parameters between two point patterns[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence, 1991, 13 (4): 376- 380. doi: 10.1109/34.88573

序列	PAN-SLAM		本文
序列	追踪数量	中断	追踪数量	中断
00	234	3	302	0
01	254	2	328	0

序列	PAN-SLAM		本文算法
序列	RMSE	STD	RMSE	STD
00	4.47	2.88	3.34	1.46
01	2.74	1.75	1.37	0.78

序列	PAN-ORBSLAM3			本文算法
序列	尺度误差/%	RMSE/m	STD/m	尺度误差/%	RMSE/m	STD/m
02	20.15	0.0166	0.0112	19.96	0.0085	0.0051
03	24.47	0.0062	0.0039	5.53	0.0046	0.0026
04	36.80	0.0045	0.0021	11.73	0.0039	0.0016
05	31.13	0.0032	0.0020	3.53	0.0020	0.0012
均值	28.14	0.0076	0.0048	10.19	0.0048	0.0026

算法	融合模式	相机数目	惯导数目
VINS-Mono	单目惯性	1	1
ORB-SLAM3	单目惯性	1	1
OpenVINS	单目惯性	1	1
OpenVINS（multi）	多组单目惯性	4	1
本文算法	全景惯性	4	1

序列	VINS-Mono		ORB-SLAM3		OpenVINS		OpenVINS（multi）		本文算法
序列	RMSE	STD	RMSE	STD	RMSE	STD	RMSE	STD	RMSE	STD
02	17.90	9.39	—	—	28.29	8.43	11.09	7.20	0.48	0.17
03	10.73	4.92	30.61	12.94	12.44	6.11	11.18	4.95	2.93	1.63
04	22.87	9.47	—	—	8.05	3.19	6.74	3.58	1.84	0.81
05	10.64	6.19	15.53	6.76	16.51	5.94	6.09	3.27	2.99	1.31