Scale Drift-Aware Large Scale Monocular SLAM

By Hauke Strasdat, José M. M. Montiel, Andrew J. Davison. Robotics: Science and Systems Conference, 2010

576 citations recorded by the source corpus.

State of the art visual SLAM systems have recently been presented which are capable of accurate, large-scale and real-time performance, but most of these require stereo vision. Important application areas in robotics and beyond open up if similar performance can be demonstrated using monocular vision, since a single camera will always be cheaper, more compact and easier to calibrate than a multi-camera rig. With high quality estimation, a single camera moving through a static scene of course effectively provides its own stereo geometry via frames distributed over time. However, a classic issue with monocular visual SLAM is that due to the purely projective nature of a single camera, motion estimates and map structure can only be recovered up to scale. Without the known inter-camera distance of a stereo rig to serve as an anchor, the scale of locally constructed map portions and the corresponding motion estimates is therefore liable to drift over time. In this paper we describe a new near real-time visual SLAM system which adopts the continuous keyframe optimisation approach of the best current stereo systems, but accounts for the additional challenges presented by monocular input. In particular, we present a new pose-graph optimisation technique which allows for the efficient correction of rotation, translation and scale drift at loop closures. Especially, we describe the Lie group of similarity transformations and its relation to the corresponding Lie algebra. We also present in detail the system’s new image processing front-end which is able accurately to track hundreds of features per frame, and a filter-based approach for feature initialisation within keyframe-based SLAM. Our approach is proven via large-scale simulation and real-world experiments where a camera completes large looped trajectories.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/17f7feb54c890a0d78df9c0a7a5c2b072b8348b4

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

ORB-SLAM3: An Accurate Open-Source Library for Visual, Visual–Inertial, and Multimap SLAM (2020, methodology): “Large scale monocular SLAM was achieved in [57] using sliding-window BA, and in [58] using a double-window optimization and a covisibility graph.”
A benchmark for the evaluation of RGB-D SLAM systems (2012, background): “Different sensor modalities have been explored in the past, including 2D laser scanners [17], [18], 3D scanners [19]– [21], monocular cameras [9], [14]–[16], [22]–[24], stereo systems [25], [26] and recently RGB-D sensors such as the Microsoft Kinect [2]–[4].”
Real-time visual odometry from dense RGB-D images (2011, background): “In contrast, many state-of-the-art approaches using monocular camera images extract key-points and match them with previous frames [3] [9] [2] using a sequence of processing steps including descriptor matching, RANSAC and bundle adjustment.”
Visual SLAM algorithms: a survey from 2010 to 2016 (2017, methodology): “[37] proposed a method for optimizing 7 DoF camera poses based on similarity transformation.”
G2o: A general framework for graph optimization (2011, methodology): “Compared to RobotVision, g 2 o is on average two times faster.”
Computer Vision for Autonomous Vehicles: Problems, Datasets and State-of-the-Art (2017, methodology): “They combine methods from loop detection (Gálvez-López & Tardós (2012)), loop closing (Strasdat et al. (2010, 2011)) and pose graph optimization (Kümmerle et al. (2011)) into one system.”
An evaluation of the RGB-D SLAM system (2012, background): “Visual SLAM systems [5], [16], [28] – in the computer vision literature often referred to as structure and motion estimation [14], [21] – typically extract sparse keypoints from the camera images.”
Large-scale direct SLAM with stereo cameras (2015, background): “More recently, Strasdat et al. [5] included scale as a parameter in a key-frame-based optimization approach to sparse monocular SLAM.”
Stereo DSO: Large-Scale Direct Sparse Visual Odometry with Stereo Cameras (2017, background): “proposed to expand the concept of keyframes to integrate scale [26] and proposed a double window optimization [25].”
ORB-SLAM2: An Open-Source SLAM System for Monocular, Stereo, and RGB-D Cameras (2016, background): “In contrast to monocular ORB-SLAM, where scale drift may occur [20], the stereo/depth information makes scale observable and the geometric validation and pose-graph optimization no longer require dealing with scale drift and are based on rigid body transformations instead of similarities.”
LSD-SLAM: Large-Scale Direct Monocular SLAM (2014, methodology): “For LSD-SLAM, we also show the number of keyframes created.”
Integrating generic sensor fusion algorithms with sound state representations through encapsulation of manifolds (2011, methodology): “[7,42,39] is to embed this idea into a mathematical and algorithmic framework by means of an explicit axiomatization, where the operators and make it easy to adapt algorithms operating on R.”
Towards a benchmark for RGB-D SLAM evaluation (2011, unknown): “Different sensor modalities have been explored in the past, including 2D laser scanners [12], [3], 3D scanners [14], [16], monocular cameras [13], [7], [9], [19], [20] and stereo systems [8].”
3-D Mapping With an RGB-D Camera (2014, methodology): “Visual SLAM approaches [8]–[10], also referred to as " structure and motion estimation " [11], [12], compute the robot's motion and the map using cameras as sensors.”
Vision-based state estimation for autonomous rotorcraft MAVs in complex environments (2013, methodology): “To this end, we are currently investigating appearance-based loop closure detection methods such as those in [13, 14].”
Visual SLAM: Why filter? (2012, unknown): “Similarly, several techniques were introduced to reduce the computational complexity of real-time BA using segment-wise optimisation [31], feature marginalisation followed by pose-graph optimisation [29, 49], incremental smoothing [26] or relative/topological representations [46].”
Deep Virtual Stereo Odometry: Leveraging Deep Depth Prediction for Monocular Direct Sparse Odometry (2018, methodology): “When using purely geometric cues, scale drift is one of the main sources of error of monocular VO due to scale unobservability [37].”