ORB-SLAM3: An Accurate Open-Source Library for Visual, Visual–Inertial, and Multimap SLAM
ORB-SLAM3: точная библиотека с открытым исходным кодом для визуального, визуально-инерциального SLAM и SLAM с использованием нескольких карт
2021-05-25
SCID: 54.1/hbxhmryb
Discuss with AI
ORB-SLAM3maximum a posteriori estimationmultimap SLAMplace recognitionvisual-inertial SLAM
Figures from the paper
Abstract (AI)
This article presents ORB-SLAM3, the first system able to perform visual, visual-inertial and multimap SLAM with monocular, stereo and RGB-D cameras, using pin-hole and fisheye lens models. The first main novelty is a tightly integrated visual-inertial SLAM system that fully relies on maximuma posteriori(MAP) estimation, even during IMU initialization, resulting in real-time robust operation in small and large, indoor and outdoor environments, being two to ten times more accurate than previous approaches. The second main novelty is a multiple map system relying on a new place recognition method with improved recall that lets ORB-SLAM3 survive to long periods of poor visual information: when it gets lost, it starts a new map that will be seamlessly merged with previous maps when revisiting them. Compared with visual odometry systems that only use information from the last few seconds, ORB-SLAM3 is the first system able to reuse in all the algorithm stages all previous information from high parallax co-visible keyframes, even if they are widely separated in time or come from previous mapping sessions, boosting accuracy. Our experiments show that, in all sensor configurations, ORB-SLAM3 is as robust as the best systems available in the literature and significantly more accurate. Notably, our stereo-inertial SLAM achieves an average accuracy of 3.5 cm in the EuRoC drone and 9 mm under quick hand-held motions in the room of TUM-VI dataset, representative of AR/VR scenarios. For the benefit of the community we make public the source code.
Key Findings
1
Across sensor configurations, ORB-SLAM3 matches the robustness of leading systems while achieving significantly higher accuracy, including 3.5 cm on EuRoC and 9 mm on rapid handheld TUM-VI sequences.
2
Its tightly integrated MAP-based visual-inertial formulation, including IMU initialization, enables robust real-time operation and achieves two- to tenfold higher accuracy than previous approaches.
3
ORB-SLAM3 is the first unified system supporting visual, visual-inertial, and multimap SLAM with monocular, stereo, and RGB-D cameras using pinhole and fisheye models.
4
ORB-SLAM3 reuses high-parallax covisible keyframes from all previous mapping sessions and widely separated times, improving accuracy over short-window visual odometry.
5
The multimap system uses improved-recall place recognition to recover from prolonged poor visual information and seamlessly merge newly created maps upon revisiting known areas.
Research Object
ORB-SLAM3 visual, visual-inertial, and multimap SLAM system operating with monocular, stereo, and RGB-D cameras
Research Subject
SLAM accuracy, robustness, real-time operation, and reuse and seamless merging of maps under varying sensor configurations and periods of poor visual information
Publication Details
Publication Date
2021-05-25
Journal
Publisher
ISSN
Cited by
5008
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest