Real-time stereo vision for collision detection on autonomous UAVs

Aleksandar Stanoev; Nicolas Audinet; Scott Tancock; Naim Dahnoun

doi:10.1109/IST.2017.8261524

Real-time stereo vision for collision detection on autonomous UAVs

Aleksandar Stanoev, Nicolas Audinet, Scott Tancock, Naim Dahnoun

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

9 Citations (Scopus)

765 Downloads (Pure)

Abstract

Collision detection is an important unsolved problem in the domain of modern UAV, which would enable safe navigation in unknown environments. Stereo vision provides a compact, lightweight and low-power solution. This paper describes an adaptive system for achieving real-time stereo vision for collision detection on an embedded GPU. Several optimisations are described including using sensor fusion with an ultrasonic sensor to better filter noise, organising the computations to take advantage of the platform's heterogeneous architecture and using GPU textures to benefit from caching. A discussion of the hardware features is provided, followed by the algorithm and implementation details for disparity calculations and finally a method for identifying objects from a disparity map. The system was implemented on an NVIDIA Tegra X1, achieving 48 FPS on a 320×240 image.

Original language	English
Title of host publication	Proceedings of the IEEE International Conference on Imaging Systems and Techniques (IST), 2017
Publisher	Institute of Electrical and Electronics Engineers (IEEE)
Number of pages	6
ISBN (Electronic)	9781538616208
ISBN (Print)	9781538616215
DOIs	https://doi.org/10.1109/IST.2017.8261524
Publication status	Published - 18 Jan 2018
Event	IEEE International Conference on Imaging Systems & Techniques - Beijing, China Duration: 18 Oct 2017 → 20 Oct 2017

Conference

Conference	IEEE International Conference on Imaging Systems & Techniques
Country/Territory	China
City	Beijing
Period	18/10/17 → 20/10/17

Research Groups and Themes

Photonics and Quantum

Access to Document

10.1109/IST.2017.8261524

Full-text PDF (accepted author manuscript)
This is the author accepted manuscript (AAM). The final published version (version of record) is available online via IEEE at http://ieeexplore.ieee.org/document/8261524/. Please refer to any applicable terms of use of the publisher.
Accepted author manuscript, 4.97 MB

Handle.net

Persistent link

Cite this

@inproceedings{8c255814f0fa416c90c57c0ad9f75834,

title = "Real-time stereo vision for collision detection on autonomous UAVs",

abstract = "Collision detection is an important unsolved problem in the domain of modern UAV, which would enable safe navigation in unknown environments. Stereo vision provides a compact, lightweight and low-power solution. This paper describes an adaptive system for achieving real-time stereo vision for collision detection on an embedded GPU. Several optimisations are described including using sensor fusion with an ultrasonic sensor to better filter noise, organising the computations to take advantage of the platform's heterogeneous architecture and using GPU textures to benefit from caching. A discussion of the hardware features is provided, followed by the algorithm and implementation details for disparity calculations and finally a method for identifying objects from a disparity map. The system was implemented on an NVIDIA Tegra X1, achieving 48 FPS on a 320×240 image.",

author = "Aleksandar Stanoev and Nicolas Audinet and Scott Tancock and Naim Dahnoun",

year = "2018",

month = jan,

day = "18",

doi = "10.1109/IST.2017.8261524",

language = "English",

isbn = "9781538616215",

booktitle = "Proceedings of the IEEE International Conference on Imaging Systems and Techniques (IST), 2017",

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

address = "United States",

note = "IEEE International Conference on Imaging Systems & Techniques ; Conference date: 18-10-2017 Through 20-10-2017",

}

Stanoev, A, Audinet, N, Tancock, S & Dahnoun, N 2018, Real-time stereo vision for collision detection on autonomous UAVs. in Proceedings of the IEEE International Conference on Imaging Systems and Techniques (IST), 2017. Institute of Electrical and Electronics Engineers (IEEE), IEEE International Conference on Imaging Systems & Techniques, Beijing, China, 18/10/17. https://doi.org/10.1109/IST.2017.8261524

Real-time stereo vision for collision detection on autonomous UAVs. / Stanoev, Aleksandar; Audinet, Nicolas; Tancock, Scott et al.
Proceedings of the IEEE International Conference on Imaging Systems and Techniques (IST), 2017. Institute of Electrical and Electronics Engineers (IEEE), 2018.

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

TY - GEN

T1 - Real-time stereo vision for collision detection on autonomous UAVs

AU - Stanoev, Aleksandar

AU - Audinet, Nicolas

AU - Tancock, Scott

AU - Dahnoun, Naim

PY - 2018/1/18

Y1 - 2018/1/18

N2 - Collision detection is an important unsolved problem in the domain of modern UAV, which would enable safe navigation in unknown environments. Stereo vision provides a compact, lightweight and low-power solution. This paper describes an adaptive system for achieving real-time stereo vision for collision detection on an embedded GPU. Several optimisations are described including using sensor fusion with an ultrasonic sensor to better filter noise, organising the computations to take advantage of the platform's heterogeneous architecture and using GPU textures to benefit from caching. A discussion of the hardware features is provided, followed by the algorithm and implementation details for disparity calculations and finally a method for identifying objects from a disparity map. The system was implemented on an NVIDIA Tegra X1, achieving 48 FPS on a 320×240 image.

AB - Collision detection is an important unsolved problem in the domain of modern UAV, which would enable safe navigation in unknown environments. Stereo vision provides a compact, lightweight and low-power solution. This paper describes an adaptive system for achieving real-time stereo vision for collision detection on an embedded GPU. Several optimisations are described including using sensor fusion with an ultrasonic sensor to better filter noise, organising the computations to take advantage of the platform's heterogeneous architecture and using GPU textures to benefit from caching. A discussion of the hardware features is provided, followed by the algorithm and implementation details for disparity calculations and finally a method for identifying objects from a disparity map. The system was implemented on an NVIDIA Tegra X1, achieving 48 FPS on a 320×240 image.

U2 - 10.1109/IST.2017.8261524

DO - 10.1109/IST.2017.8261524

M3 - Conference Contribution (Conference Proceeding)

SN - 9781538616215

BT - Proceedings of the IEEE International Conference on Imaging Systems and Techniques (IST), 2017

PB - Institute of Electrical and Electronics Engineers (IEEE)

T2 - IEEE International Conference on Imaging Systems & Techniques

Y2 - 18 October 2017 through 20 October 2017

ER -

Real-time stereo vision for collision detection on autonomous UAVs

Abstract

Conference

Research Groups and Themes

Access to Document

Handle.net

Fingerprint

Cite this