Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

IDLS: Inverse Depth Line based Visual-Inertial SLAM

Wanting Li, Shuo Wang, Yongcai Wang, Yu Shao, Xuewei Bai, Deying Li

TL;DR

This work addresses robust monocular visual-inertial SLAM in indoor environments by exploiting line geometry. It introduces an inverse-depth line representation parameterized by the two endpoints’ depths, enabling a compact 2-DOF line state and efficient multi-frame triangulation. A robust line triangulation method and a line reprojection error model based on the line normal vector are combined with a two-step optimization to alternately refine line endpoints and camera poses, reducing computational burden. Integrated into a three-thread IDLS system, the approach outperforms state-of-the-art point-line SLAM methods on EuRoC and real indoor datasets in terms of both accuracy and efficiency. The results demonstrate improved robustness in texture-poor settings and offer practical benefits for real-time navigation and mapping.

Abstract

For robust visual-inertial SLAM in perceptually-challenging indoor environments,recent studies exploit line features to extract descriptive information about scene structure to deal with the degeneracy of point features. But existing point-line-based SLAM methods mainly use Plücker matrix or orthogonal representation to represent a line, which needs to calculate at least four variables to determine a line. Given the numerous line features to determine in each frame, the overly flexible line representation increases the computation burden and comprises the accuracy of the results. In this paper, we propose inverse depth representation for a line, which models each extracted line feature using only two variables, i.e., the inverse depths of the two ending points. It exploits the fact that the projected line's pixel coordinates on the image plane are rather accurate, which partially restrict the line. Using this compact line presentation, Inverse Depth Line SLAM (IDLS) is proposed to track the line features in SLAM in an accurate and efficient way. A robust line triangulation method and a novel line re-projection error model are introduced. And a two-step optimization method is proposed to firstly determine the lines and then to estimate the camera poses in each frame. IDLS is extensively evaluated in multiple perceptually-challenging datasets. The results show it is more accurate, robust, and needs lower computational overhead than the current state-of-the-art of point-line-based SLAM methods.

IDLS: Inverse Depth Line based Visual-Inertial SLAM

TL;DR

This work addresses robust monocular visual-inertial SLAM in indoor environments by exploiting line geometry. It introduces an inverse-depth line representation parameterized by the two endpoints’ depths, enabling a compact 2-DOF line state and efficient multi-frame triangulation. A robust line triangulation method and a line reprojection error model based on the line normal vector are combined with a two-step optimization to alternately refine line endpoints and camera poses, reducing computational burden. Integrated into a three-thread IDLS system, the approach outperforms state-of-the-art point-line SLAM methods on EuRoC and real indoor datasets in terms of both accuracy and efficiency. The results demonstrate improved robustness in texture-poor settings and offer practical benefits for real-time navigation and mapping.

Abstract

For robust visual-inertial SLAM in perceptually-challenging indoor environments,recent studies exploit line features to extract descriptive information about scene structure to deal with the degeneracy of point features. But existing point-line-based SLAM methods mainly use Plücker matrix or orthogonal representation to represent a line, which needs to calculate at least four variables to determine a line. Given the numerous line features to determine in each frame, the overly flexible line representation increases the computation burden and comprises the accuracy of the results. In this paper, we propose inverse depth representation for a line, which models each extracted line feature using only two variables, i.e., the inverse depths of the two ending points. It exploits the fact that the projected line's pixel coordinates on the image plane are rather accurate, which partially restrict the line. Using this compact line presentation, Inverse Depth Line SLAM (IDLS) is proposed to track the line features in SLAM in an accurate and efficient way. A robust line triangulation method and a novel line re-projection error model are introduced. And a two-step optimization method is proposed to firstly determine the lines and then to estimate the camera poses in each frame. IDLS is extensively evaluated in multiple perceptually-challenging datasets. The results show it is more accurate, robust, and needs lower computational overhead than the current state-of-the-art of point-line-based SLAM methods.
Paper Structure (29 sections, 2 theorems, 17 equations, 9 figures, 7 tables, 1 algorithm)

This paper contains 29 sections, 2 theorems, 17 equations, 9 figures, 7 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Front-end Approaches
  4. Line Extraction
  5. Line Tracking
  6. Line Triangulation
  7. Back-end Approaches
  8. Loop Closure
  9. Problem Model and Inverse Depth Line Geometry
  10. Line Feature Representation
  11. Plücker Coordinate Representation
  12. Orthonormal Representation
  13. Inverse Depth Representation
  14. Line's Reprojection Residual Error Model
  15. Inverse Depth Line State Initialization
  16. ...and 14 more sections

Key Result

Proposition 1

In calculating the frame-to-frame line re-projection residual error using the matched line segments $L_{c_1}$ and $L_{c_2}$ captured by $c_1$ and $c_2$ respectively, these two captured line segments don't need to have common 3D ending vertices.

Figures (9)

  • Figure 1: DOFs of different line representations.
  • Figure 2: Figure shows the IDLS's trajectory and point-line map for the MH-04-Difficult sequence. The two images are screenshots of the ROS Rviz window, where the red lines represent tracked lines, the blue lines represent untracked lines, and the green line represents the motion trajectory. (a) SLAM trajectory and feature map window. (b) Feature tracked window.
  • Figure 3: The line reprojection residual of an observed line in a sliding window is modelled in terms of the point-to-line distance.
  • Figure 4: Initialization of a newly observed line. The straight line $L_{c_1}$ can be regarded as the intersection of two planes $\pi_1$ and $\pi _2$ to obtain, that is, this line is in both planes and the distance to both planes is 0.
  • Figure 5: System overview of IDLS. It consists of three threads: measurement preprocessing, VIO, and loop closure.
  • ...and 4 more figures

Theorems & Definitions (2)

  • Proposition 1
  • Proposition 2