Image Enhancement using GANs for Monocular Visual Odometry

Jon Zubieta Ansorregi; Mikel Etxeberria Garcia; Maider Zamalloa Akizu; Nestor Arana Arexolaleiba

doi:10.1109/ECMSM51310.2021.9468831

Image Enhancement using GANs for Monocular Visual Odometry

Jon Zubieta Ansorregi, Mikel Etxeberria Garcia, Maider Zamalloa Akizu, Nestor Arana Arexolaleiba

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

1 Citation (Scopus)

Abstract

Drones, mobile robots, and autonomous vehicles use Visual Odometry (VO) to move around complex environments. ORB-SLAM or deep learning-based approaches like DF-VO are two of the state-of-the-art technics for monocular VO. Those two technics perform correctly in outdoor scenarios but show some limitations in indoor environments. The extreme lighting conditions, non-Lambertian surfaces, or occlusion of indoor environments can disturb the visual information, and so the odometry information. Generative Adversarial Network (GAN) architectures recently proposed in the literature can help to overcome image low-light and blurring limitations. This research study aims to assess image enhancement's impact using GANS on the Visual Odometry algorithm DF-VO. Since DF-VO is also based on visual geometric information, the paper first considers the effect of two different GAN architectures in the camera's calibration. Then, the impact in the odometry information computed by DF-VO is evaluated. The preliminary results show that the reprojection error and the uncertainty of the calibration of a pin-hole-based camera do not increase significantly, and DF-VO's performance is improved.

Original language	English
Title of host publication	2021 IEEE International Workshop of Electronics, Control, Measurement, Signals and their Application to Mechatronics, ECMSM 2021
Publisher	IEEE
Publication date	21 Jun 2021
Article number	9468831
ISBN (Print)	978-1-5386-1796-0
ISBN (Electronic)	9781538617571
DOIs	https://doi.org/10.1109/ECMSM51310.2021.9468831
Publication status	Published - 21 Jun 2021
Event	15th IEEE International Workshop of Electronics, Control, Measurement, Signals and their application to Mechatronics, ECMSM 2021 - Liberec, Czech Republic Duration: 21 Jun 2021 → 22 Jun 2021

Conference

Conference	15th IEEE International Workshop of Electronics, Control, Measurement, Signals and their application to Mechatronics, ECMSM 2021
Country/Territory	Czech Republic
City	Liberec
Period	21/06/2021 → 22/06/2021
Sponsor	IAS - IEEE Industrial Application Society, IEEE

Bibliographical note

Funding Information:
This work was supported in part by the Basque Government through BIKAINTEK2018 program and CAF Signalling. The authors would also like to acknowledge support by European Union s H2020-WIDESPREAD project no. 857061 "Networking for Research and Development of Human Interactive and Sensitive Robotics Taking Advantage of Additive Manufacturing - R2P2". Finally, the authors will thanks to Daniel Maestro Watson for his input on uncertainties.

Funding Information:
ACKNOWLEDGMENT This work was supported in part by the Basque Government through BIKAINTEK2018 program and CAF Signalling. The authors would also like to acknowledge support by European Union’s H2020-WIDESPREAD project no. 857061 “Networking for Research and Development of Human Interactive and Sensitive Robotics Taking Advantage of Additive Manufacturing - R2P2”. Finally, the authors will thanks to Daniel Maestro Watson for his input on uncertainties.

Publisher Copyright:
© 2021 IEEE.

Keywords

Calibration
Deep Learning
Image enhancement
Visual Odometry

Access to Document

10.1109/ECMSM51310.2021.9468831

AUB Link

Search for the material in Aalborg University Library's search engine

Cite this

@inproceedings{157d2b27dbd447a0b86bd04221431162,

title = "Image Enhancement using GANs for Monocular Visual Odometry",

abstract = "Drones, mobile robots, and autonomous vehicles use Visual Odometry (VO) to move around complex environments. ORB-SLAM or deep learning-based approaches like DF-VO are two of the state-of-the-art technics for monocular VO. Those two technics perform correctly in outdoor scenarios but show some limitations in indoor environments. The extreme lighting conditions, non-Lambertian surfaces, or occlusion of indoor environments can disturb the visual information, and so the odometry information. Generative Adversarial Network (GAN) architectures recently proposed in the literature can help to overcome image low-light and blurring limitations. This research study aims to assess image enhancement's impact using GANS on the Visual Odometry algorithm DF-VO. Since DF-VO is also based on visual geometric information, the paper first considers the effect of two different GAN architectures in the camera's calibration. Then, the impact in the odometry information computed by DF-VO is evaluated. The preliminary results show that the reprojection error and the uncertainty of the calibration of a pin-hole-based camera do not increase significantly, and DF-VO's performance is improved.",

keywords = "Calibration, Deep Learning, Image enhancement, Visual Odometry",

author = "Ansorregi, {Jon Zubieta} and Garcia, {Mikel Etxeberria} and Akizu, {Maider Zamalloa} and Arexolaleiba, {Nestor Arana}",

note = "Funding Information: This work was supported in part by the Basque Government through BIKAINTEK2018 program and CAF Signalling. The authors would also like to acknowledge support by European Union s H2020-WIDESPREAD project no. 857061 {"}Networking for Research and Development of Human Interactive and Sensitive Robotics Taking Advantage of Additive Manufacturing - R2P2{"}. Finally, the authors will thanks to Daniel Maestro Watson for his input on uncertainties. Funding Information: ACKNOWLEDGMENT This work was supported in part by the Basque Government through BIKAINTEK2018 program and CAF Signalling. The authors would also like to acknowledge support by European Union{\textquoteright}s H2020-WIDESPREAD project no. 857061 “Networking for Research and Development of Human Interactive and Sensitive Robotics Taking Advantage of Additive Manufacturing - R2P2”. Finally, the authors will thanks to Daniel Maestro Watson for his input on uncertainties. Publisher Copyright: {\textcopyright} 2021 IEEE.; 15th IEEE International Workshop of Electronics, Control, Measurement, Signals and their application to Mechatronics, ECMSM 2021 ; Conference date: 21-06-2021 Through 22-06-2021",

year = "2021",

month = jun,

day = "21",

doi = "10.1109/ECMSM51310.2021.9468831",

language = "English",

isbn = "978-1-5386-1796-0",

booktitle = "2021 IEEE International Workshop of Electronics, Control, Measurement, Signals and their Application to Mechatronics, ECMSM 2021",

publisher = "IEEE",

address = "United States",

}

Ansorregi, JZ, Garcia, ME, Akizu, MZ & Arexolaleiba, NA 2021, Image Enhancement using GANs for Monocular Visual Odometry. in 2021 IEEE International Workshop of Electronics, Control, Measurement, Signals and their Application to Mechatronics, ECMSM 2021., 9468831, IEEE, 15th IEEE International Workshop of Electronics, Control, Measurement, Signals and their application to Mechatronics, ECMSM 2021, Liberec, Czech Republic, 21/06/2021. https://doi.org/10.1109/ECMSM51310.2021.9468831

Image Enhancement using GANs for Monocular Visual Odometry. / Ansorregi, Jon Zubieta; Garcia, Mikel Etxeberria; Akizu, Maider Zamalloa et al.
2021 IEEE International Workshop of Electronics, Control, Measurement, Signals and their Application to Mechatronics, ECMSM 2021. IEEE, 2021. 9468831.

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

TY - GEN

T1 - Image Enhancement using GANs for Monocular Visual Odometry

AU - Ansorregi, Jon Zubieta

AU - Garcia, Mikel Etxeberria

AU - Akizu, Maider Zamalloa

AU - Arexolaleiba, Nestor Arana

N1 - Funding Information: This work was supported in part by the Basque Government through BIKAINTEK2018 program and CAF Signalling. The authors would also like to acknowledge support by European Union s H2020-WIDESPREAD project no. 857061 "Networking for Research and Development of Human Interactive and Sensitive Robotics Taking Advantage of Additive Manufacturing - R2P2". Finally, the authors will thanks to Daniel Maestro Watson for his input on uncertainties. Funding Information: ACKNOWLEDGMENT This work was supported in part by the Basque Government through BIKAINTEK2018 program and CAF Signalling. The authors would also like to acknowledge support by European Union’s H2020-WIDESPREAD project no. 857061 “Networking for Research and Development of Human Interactive and Sensitive Robotics Taking Advantage of Additive Manufacturing - R2P2”. Finally, the authors will thanks to Daniel Maestro Watson for his input on uncertainties. Publisher Copyright: © 2021 IEEE.

PY - 2021/6/21

Y1 - 2021/6/21

N2 - Drones, mobile robots, and autonomous vehicles use Visual Odometry (VO) to move around complex environments. ORB-SLAM or deep learning-based approaches like DF-VO are two of the state-of-the-art technics for monocular VO. Those two technics perform correctly in outdoor scenarios but show some limitations in indoor environments. The extreme lighting conditions, non-Lambertian surfaces, or occlusion of indoor environments can disturb the visual information, and so the odometry information. Generative Adversarial Network (GAN) architectures recently proposed in the literature can help to overcome image low-light and blurring limitations. This research study aims to assess image enhancement's impact using GANS on the Visual Odometry algorithm DF-VO. Since DF-VO is also based on visual geometric information, the paper first considers the effect of two different GAN architectures in the camera's calibration. Then, the impact in the odometry information computed by DF-VO is evaluated. The preliminary results show that the reprojection error and the uncertainty of the calibration of a pin-hole-based camera do not increase significantly, and DF-VO's performance is improved.

AB - Drones, mobile robots, and autonomous vehicles use Visual Odometry (VO) to move around complex environments. ORB-SLAM or deep learning-based approaches like DF-VO are two of the state-of-the-art technics for monocular VO. Those two technics perform correctly in outdoor scenarios but show some limitations in indoor environments. The extreme lighting conditions, non-Lambertian surfaces, or occlusion of indoor environments can disturb the visual information, and so the odometry information. Generative Adversarial Network (GAN) architectures recently proposed in the literature can help to overcome image low-light and blurring limitations. This research study aims to assess image enhancement's impact using GANS on the Visual Odometry algorithm DF-VO. Since DF-VO is also based on visual geometric information, the paper first considers the effect of two different GAN architectures in the camera's calibration. Then, the impact in the odometry information computed by DF-VO is evaluated. The preliminary results show that the reprojection error and the uncertainty of the calibration of a pin-hole-based camera do not increase significantly, and DF-VO's performance is improved.

KW - Calibration

KW - Deep Learning

KW - Image enhancement

KW - Visual Odometry

UR - http://www.scopus.com/inward/record.url?scp=85112564759&partnerID=8YFLogxK

U2 - 10.1109/ECMSM51310.2021.9468831

DO - 10.1109/ECMSM51310.2021.9468831

M3 - Article in proceeding

AN - SCOPUS:85112564759

SN - 978-1-5386-1796-0

BT - 2021 IEEE International Workshop of Electronics, Control, Measurement, Signals and their Application to Mechatronics, ECMSM 2021

PB - IEEE

T2 - 15th IEEE International Workshop of Electronics, Control, Measurement, Signals and their application to Mechatronics, ECMSM 2021

Y2 - 21 June 2021 through 22 June 2021

ER -

Image Enhancement using GANs for Monocular Visual Odometry

Abstract

Conference

Bibliographical note

Keywords

Access to Document

AUB Link

Other files and links

Fingerprint

Cite this