Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Inline Photometrically Calibrated Hybrid Visual SLAM

Nicolas Abboud, Malak Sayour, Imad H. Elhajj, John Zelek, Daniel Asmar

TL;DR

This paper presents an integrated approach to Visual SLAM, merging online sequential photometric calibration within a Hybrid direct-indirect visual SLAM (H-SLAM) that outperforms other state of the art direct, indirect, and hybrid Visual SLAM systems.

Abstract

This paper presents an integrated approach to Visual SLAM, merging online sequential photometric calibration within a Hybrid direct-indirect visual SLAM (H-SLAM). Photometric calibration helps normalize pixel intensity values under different lighting conditions, and thereby improves the direct component of our H-SLAM. A tangential benefit also results to the indirect component of H-SLAM given that the detected features are more stable across variable lighting conditions. Our proposed photometrically calibrated H-SLAM is tested on several datasets, including the TUM monoVO as well as on a dataset we created. Calibrated H-SLAM outperforms other state of the art direct, indirect, and hybrid Visual SLAM systems in all the experiments. Furthermore, in online SLAM tested at our site, it also significantly outperformed the other SLAM Systems.

Inline Photometrically Calibrated Hybrid Visual SLAM

TL;DR

This paper presents an integrated approach to Visual SLAM, merging online sequential photometric calibration within a Hybrid direct-indirect visual SLAM (H-SLAM) that outperforms other state of the art direct, indirect, and hybrid Visual SLAM systems.

Abstract

This paper presents an integrated approach to Visual SLAM, merging online sequential photometric calibration within a Hybrid direct-indirect visual SLAM (H-SLAM). Photometric calibration helps normalize pixel intensity values under different lighting conditions, and thereby improves the direct component of our H-SLAM. A tangential benefit also results to the indirect component of H-SLAM given that the detected features are more stable across variable lighting conditions. Our proposed photometrically calibrated H-SLAM is tested on several datasets, including the TUM monoVO as well as on a dataset we created. Calibrated H-SLAM outperforms other state of the art direct, indirect, and hybrid Visual SLAM systems in all the experiments. Furthermore, in online SLAM tested at our site, it also significantly outperformed the other SLAM Systems.
Paper Structure (17 sections, 5 equations, 6 figures)

This paper contains 17 sections, 5 equations, 6 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Hybrid Visual SLAM: Bridging Direct and Indirect Methods
  4. Photometric Calibration: Enhancing Image Understanding
  5. Photometrically Calibrated Hybrid Visual SLAM
  6. Proposed System
  7. Descriptor sharing and feature extraction interplay
  8. Sequential in-line photometric calibration
  9. Harmonized integration: joint optimization, map maintenance and loop closure enhancement
  10. Experiments
  11. Dataset Evaluation
  12. Evaluation in diverse operational settings
  13. Live real-time evaluation
  14. Results and discussion
  15. Dataset observations
  16. ...and 2 more sections

Figures (6)

  • Figure 1: Sample output of the photmetrically calibrated hybrid SLAM system run on Sequence 30 of the TUM monoVO dataset.
  • Figure 2: Diagram of the proposed system showing the integration of sequential photometric calibration with the multi-threaded architecture of Hybrid SLAM
  • Figure 3: Full evaluation results, showing cumulative alignment error for all tested sequences in the TUM-monoVO dataset. Each square corresponds to the color-coded alignment error, as defined in engel2016photometrically. We run each of the tested sequence (horizontal axis) 5 times each (vertical axis).
  • Figure 4: Accumulated alignment error, rotational drift error and translational error as defined in engel2016photometrically for each system over all runs
  • Figure 5: Tabulated results of the live experiment conducted on campus at the American University of Beirut. The table shows the trajectories of HSLAM and DSO under two calibration modes alongside the "ground truth" obtained from RTK-GPS
  • ...and 1 more figures