From Christian Forster via ros-users@
Dear Colleagues,
We would like to draw your attention to a new open-source monocular
visual odometry algorithm called SVO (``Semi-direct Visual Odometry'').
The semi-direct approach eliminates the need of costly feature
extraction and robust matching techniques for motion estimation. Our
algorithm operates directly on pixel intensities, which results in
subpixel precision at high frame-rates (up to 70 fps on latest
smartphone processors, up to 400 fps on laptops). A probabilistic
mapping method that explicitly models outlier measurements is used to
estimate 3D points, which results in fewer outliers and more reliable
points. Precise and high frame-rate motion estimation brings increased
robustness in scenes of repetitive, and high-frequency texture.
In our group, we use SVO for vision-based, on-board ego-motion
estimation of micro aerial vehicles, which also allows them to navigate
fully autonomously.
Please check the video with a demonstration of the system capabilities:
http://youtu.be/2YnIMfw6bJY
The code is hosted on GitHub:
https://github.com/uzh-rpg/rpg_svo
A closed-source professional edition is available for commercial
purposes. Please contact us for further info.
The algorithm is described in our the paper (please cite it if you use
it for your publications):
C. Forster, M. Pizzoli, D. Scaramuzza, "SVO: Fast Semi-Direct Monocular
Visual Odometry," IEEE International Conference on Robotics and
Automation (ICRA), 2014.
PDF: http://rpg.ifi.uzh.ch/docs/ICRA14_Forster.pdf
Best regards,
Christian Forster, Matia Pizzoli, Davide Scaramuzza
We would like to draw your attention to a new open-source monocular
visual odometry algorithm called SVO (``Semi-direct Visual Odometry'').
The semi-direct approach eliminates the need of costly feature
extraction and robust matching techniques for motion estimation. Our
algorithm operates directly on pixel intensities, which results in
subpixel precision at high frame-rates (up to 70 fps on latest
smartphone processors, up to 400 fps on laptops). A probabilistic
mapping method that explicitly models outlier measurements is used to
estimate 3D points, which results in fewer outliers and more reliable
points. Precise and high frame-rate motion estimation brings increased
robustness in scenes of repetitive, and high-frequency texture.
In our group, we use SVO for vision-based, on-board ego-motion
estimation of micro aerial vehicles, which also allows them to navigate
fully autonomously.
Please check the video with a demonstration of the system capabilities:
http://youtu.be/2YnIMfw6bJY
The code is hosted on GitHub:
https://github.com/uzh-rpg/
A closed-source professional edition is available for commercial
purposes. Please contact us for further info.
The algorithm is described in our the paper (please cite it if you use
it for your publications):
C. Forster, M. Pizzoli, D. Scaramuzza, "SVO: Fast Semi-Direct Monocular
Visual Odometry," IEEE International Conference on Robotics and
Automation (ICRA), 2014.
PDF: http://rpg.ifi.uzh.ch/docs/
Best regards,
Christian Forster, Matia Pizzoli, Davide Scaramuzza
Leave a comment