HDRFusion: HDR SLAM using a low-cost auto-exposure RGB-D sensor

Shuda Li; Ankur Handa; Yang Zhang; Andrew Calway

doi:10.1109/3DV.2016.40

HDRFusion: HDR SLAM using a low-cost auto-exposure RGB-D sensor

Shuda Li
, Ankur Handa
, Yang Zhang
, Andrew Calway

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

19 Citations (Scopus)

543 Downloads (Pure)

Abstract

Most dense RGB/RGB-D SLAM systems require the brightness of 3-D points observed from different viewpoints to be constant. However, in reality, this assumption is dif- ficult to meet even when the surface is Lambertian and il- lumination is static. One cause is that most cameras auto- matically tune exposure to adapt to the wide dynamic range of scene radiance, violating the brightness assumption. We describe a novel system - HDRFusion - which turns this ap- parent drawback into an advantage by fusing LDR frames into an HDR textured volume using a standard RGB-D sen- sor with auto-exposure (AE) enabled. The key contribution is the use of a normalised metric for frame alignment which is invariant to changes in exposure time. This enables robust tracking in frame-to-model mode and also compensates the exposure accurately so that HDR texture, free of artefacts, can be generated online. We demonstrate that the track- ing robustness and accuracy is greatly improved by the ap- proach and that radiance maps can be generated with far greater dynamic range of scene radiance.

Original language	English
Title of host publication	2016 Fourth International Conference on 3D Vision (3DV 2016)
Subtitle of host publication	Proceedings of a meeting held 25-28 October 2016, Stanford, CA, USA
Publisher	Institute of Electrical and Electronics Engineers (IEEE)
Pages	314-322
Number of pages	9
ISBN (Electronic)	9781509054077
ISBN (Print)	9781509054084
DOIs	https://doi.org/10.1109/3DV.2016.40
Publication status	Published - 19 Dec 2016
Event	2016 International Conference on 3D Vision - University of Stanford, California, United States Duration: 25 Oct 2016 → 28 Oct 2016 http://3dv.stanford.edu/index.html

Conference

Conference	2016 International Conference on 3D Vision
Abbreviated title	3D Vision 2016
Country/Territory	United States
City	California
Period	25/10/16 → 28/10/16
Internet address	http://3dv.stanford.edu/index.html

Keywords

high dynamic range
3-D mapping and tracking
auto exposure
RGB-D cameras

Access to Document

10.1109/3DV.2016.40

Full-text PDF (accepted author manuscript)
This is the accepted author manuscript (AAM). The final published version (version of record) is available online via IEEE at https://doi.org/10.1109/3DV.2016.40. Please refer to any applicable terms of use of the publisher.
Accepted author manuscript, 7.31 MB

Handle.net

Persistent link

19 Citations
1 Working paper

HDRFusion: HDR SLAM using a low-cost auto-exposure RGB-D sensor
Li, S., Handa, A., Zhang, Y. & Calway, A., 4 Apr 2016, arXiv.org.
Research output: Working paper

Open Access
File

Professor Andrew Calway
- Andrew.Calwaybristol.acuk
- School of Computer Science - Professor of Computer Vision
- Visual Information Laboratory
Person: Academic , Member

Cite this

@inproceedings{84d328a6f770425482b0aafdecc5741c,

title = "HDRFusion: HDR SLAM using a low-cost auto-exposure RGB-D sensor",

abstract = "Most dense RGB/RGB-D SLAM systems require the brightness of 3-D points observed from different viewpoints to be constant. However, in reality, this assumption is dif- ficult to meet even when the surface is Lambertian and il- lumination is static. One cause is that most cameras auto- matically tune exposure to adapt to the wide dynamic range of scene radiance, violating the brightness assumption. We describe a novel system - HDRFusion - which turns this ap- parent drawback into an advantage by fusing LDR frames into an HDR textured volume using a standard RGB-D sen- sor with auto-exposure (AE) enabled. The key contribution is the use of a normalised metric for frame alignment which is invariant to changes in exposure time. This enables robust tracking in frame-to-model mode and also compensates the exposure accurately so that HDR texture, free of artefacts, can be generated online. We demonstrate that the track- ing robustness and accuracy is greatly improved by the ap- proach and that radiance maps can be generated with far greater dynamic range of scene radiance.",

keywords = "high dynamic range, 3-D mapping and tracking, auto exposure, RGB-D cameras",

author = "Shuda Li and Ankur Handa and Yang Zhang and Andrew Calway",

year = "2016",

month = dec,

day = "19",

doi = "10.1109/3DV.2016.40",

language = "English",

isbn = "9781509054084",

pages = "314--322",

booktitle = "2016 Fourth International Conference on 3D Vision (3DV 2016)",

publisher = "Institute of Electrical and Electronics Engineers (IEEE)",

address = "United States",

note = "2016 International Conference on 3D Vision, 3D Vision 2016 ; Conference date: 25-10-2016 Through 28-10-2016",

url = "http://3dv.stanford.edu/index.html",

}

Li, S, Handa, A, Zhang, Y & Calway, A 2016, HDRFusion: HDR SLAM using a low-cost auto-exposure RGB-D sensor. in 2016 Fourth International Conference on 3D Vision (3DV 2016): Proceedings of a meeting held 25-28 October 2016, Stanford, CA, USA. Institute of Electrical and Electronics Engineers (IEEE), pp. 314-322, 2016 International Conference on 3D Vision, California, United States, 25/10/16. https://doi.org/10.1109/3DV.2016.40

HDRFusion: HDR SLAM using a low-cost auto-exposure RGB-D sensor. / Li, Shuda; Handa, Ankur; Zhang, Yang et al.
2016 Fourth International Conference on 3D Vision (3DV 2016): Proceedings of a meeting held 25-28 October 2016, Stanford, CA, USA. Institute of Electrical and Electronics Engineers (IEEE), 2016. p. 314-322.

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

TY - GEN

T1 - HDRFusion

T2 - 2016 International Conference on 3D Vision

AU - Li, Shuda

AU - Handa, Ankur

AU - Zhang, Yang

AU - Calway, Andrew

PY - 2016/12/19

Y1 - 2016/12/19

N2 - Most dense RGB/RGB-D SLAM systems require the brightness of 3-D points observed from different viewpoints to be constant. However, in reality, this assumption is dif- ficult to meet even when the surface is Lambertian and il- lumination is static. One cause is that most cameras auto- matically tune exposure to adapt to the wide dynamic range of scene radiance, violating the brightness assumption. We describe a novel system - HDRFusion - which turns this ap- parent drawback into an advantage by fusing LDR frames into an HDR textured volume using a standard RGB-D sen- sor with auto-exposure (AE) enabled. The key contribution is the use of a normalised metric for frame alignment which is invariant to changes in exposure time. This enables robust tracking in frame-to-model mode and also compensates the exposure accurately so that HDR texture, free of artefacts, can be generated online. We demonstrate that the track- ing robustness and accuracy is greatly improved by the ap- proach and that radiance maps can be generated with far greater dynamic range of scene radiance.

AB - Most dense RGB/RGB-D SLAM systems require the brightness of 3-D points observed from different viewpoints to be constant. However, in reality, this assumption is dif- ficult to meet even when the surface is Lambertian and il- lumination is static. One cause is that most cameras auto- matically tune exposure to adapt to the wide dynamic range of scene radiance, violating the brightness assumption. We describe a novel system - HDRFusion - which turns this ap- parent drawback into an advantage by fusing LDR frames into an HDR textured volume using a standard RGB-D sen- sor with auto-exposure (AE) enabled. The key contribution is the use of a normalised metric for frame alignment which is invariant to changes in exposure time. This enables robust tracking in frame-to-model mode and also compensates the exposure accurately so that HDR texture, free of artefacts, can be generated online. We demonstrate that the track- ing robustness and accuracy is greatly improved by the ap- proach and that radiance maps can be generated with far greater dynamic range of scene radiance.

KW - high dynamic range

KW - 3-D mapping and tracking

KW - auto exposure

KW - RGB-D cameras

U2 - 10.1109/3DV.2016.40

DO - 10.1109/3DV.2016.40

M3 - Conference Contribution (Conference Proceeding)

SN - 9781509054084

SP - 314

EP - 322

BT - 2016 Fourth International Conference on 3D Vision (3DV 2016)

PB - Institute of Electrical and Electronics Engineers (IEEE)

Y2 - 25 October 2016 through 28 October 2016

ER -

HDRFusion: HDR SLAM using a low-cost auto-exposure RGB-D sensor

Abstract

Conference

Keywords

Access to Document

Handle.net

Fingerprint

Research output

HDRFusion: HDR SLAM using a low-cost auto-exposure RGB-D sensor

Profiles

Professor Andrew Calway

Cite this