PKS-ORB-SLAM: Photogrammetric Key-frame Selection

Authors: Arash Azimi

An improved Photogrammetric key-frame selection method built on ORB-SLAM3

To evaluate the performance of the PKS algorithm, two different datasets were used: the EuRoC Micro Aerial Vehicles (MAV) dataset as well as an in-house dataset acquired with a developed stereo system. Both datasets contain stereo images and synchronized IMU measurements. The MAV dataset have also accurate motion and structure ground-truth.

1. Summary

PKS flowchart

PKS schematic diagram

Instrument

Trajectory comparison

Point-cloud comparison

Point-cloud cross-section

Point-cloud plane fitting

Plane fitting statistics

cumulative Absolute Trajectory Error (ATE) comparison

Average ATE Results in each dataset

Key-frame trajectories

2. Related Publications

Azimi, A., Ahmadabadian, A.H. and Remondino, F., 2022. PKS: A photogrammetric key-frame selection method for visual-inertial systems built on ORB-SLAM3. ISPRS Journal of Photogrammetry and Remote Sensing, 191, pp.18-32.

Azimi, A., Hosseininaveh, A. and Remondino, F., 2022. A NOVEL GEOMETRIC KEY-FRAME SELECTION METHOD FOR VISUAL-INERTIAL SLAM AND ODOMETRY SYSTEMS. The International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, 43, pp.9-14.PDF

3. License

This repository is released under Apache License V2. To develop, publication and use it, please follow the terms of this license.

4. Citation

If you use PKS in an academic work, please cite:

@article{AZIMI202218,
  title={PKS: A photogrammetric key-frame selection method for visual-inertial systems built on ORB-SLAM3},
  author={Arash Azimi and Ali {Hosseininaveh Ahmadabadian} and Fabio Remondino},
  journal={ISPRS Journal of Photogrammetry and Remote Sensing},
  volume = {191},
  pages = {18-32},
  issn = {0924-2716},
  year={2022}
  doi={https://doi.org/10.1016/j.isprsjprs.2022.07.003}
 }

@article{
  title={a Novel Geometric Key-Frame Selection Method for Visual-Inertial Slam and Odometry Systems},
  author={Arash Azimi and Ali {Hosseininaveh Ahmadabadian} and Fabio Remondino},
  journal={The International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences},
  volume = {43},
  pages = {9-14},
  issn = {1682-1750},
  year={2022}
  doi={https://doi.org/10.5194/isprs-archives-XLIII-B2-2022-9-2022}
 }

5. Building

To build this repository, you can follow all the steps mentioned in the original version of ORB-SLAM3 in https://github.com/UZ-SLAMLab/ORB_SLAM3

If you have a problem building this repository, you can first install the original version of ORB-SLAM3 and then replace the include and src files of this repository with the original files and rebuild it.

arashazimi0032/PKS