Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Use of very wide angle fisheye cameras #131

Closed
visocopter opened this issue Jan 14, 2021 · 4 comments
Closed

Use of very wide angle fisheye cameras #131

visocopter opened this issue Jan 14, 2021 · 4 comments
Labels
question Theory or implementation question

Comments

@visocopter
Copy link

Hi,

is it possible to use a fisheye camera with a very high FOV (>180 deg) fisheye lens with Open Vins?

I have a global shutter camera and would like to use a circular 180 degree fisheye lens.
As I understood the documentation, there are two camera models supported: pinhole-equi and pinhole-radtan.

Would the pinhole-equi model sufficient for this task?
Or generally: What would have to be done to use a circular 180 degree fisheye camera?

Keep up the great work!

Cheers,
James

@goldbattle goldbattle added the question Theory or implementation question label Jan 14, 2021
@goldbattle
Copy link
Member

We leverage the opencv fisheye mode:
https://docs.opencv.org/master/db/d58/group__calib3d__fisheye.html

My understanding that their model does not support very large FOV cameras. Additionally, I think some of the estimation code for triangulating features would need to be changed as right now we assert that features must lie in front of the camera, thus for >180deg this won't hold since features can lie behind the camera frame. I don't have any experience with these cameras, but I think another issue is performing feature tracking on these large FOV angles due to the large amounts of distortion. You can try selecting a region of interest in the middle of the image to reduce the FOV is you are stuck with a high FOV sensor.

@visocopter
Copy link
Author

Alright thank you very much.
Would a 160 degree lens work then with the existing models since the image wouldn't be behind the frame?
(Like the Intel Realsense T265, which has a 160 degree FOV I think)

@goldbattle
Copy link
Member

@visocopter
Copy link
Author

Okay thank you!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
question Theory or implementation question
Projects
None yet
Development

No branches or pull requests

2 participants