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

Floor Plan Alignment for Different cameras orientation #6

Open
drochairobot opened this issue Oct 16, 2013 · 15 comments
Open

Floor Plan Alignment for Different cameras orientation #6

drochairobot opened this issue Oct 16, 2013 · 15 comments
Labels

Comments

@drochairobot
Copy link

Hi Akram,

We have a floor plan that includes 2 cameras that have an approximate orientation of 90 degrees between them. When creating the floor plan, it shows a miss-alignment of the checkerboard:
floorplanturning

This will result on tracking miss aligned. The next image highlights a trajectory that should be a straight line:
track_result_img_out

In the case we do not turn the image of the second camera when doing the floor plan, the floor plan and trajectory will appear better aligned,
track_result_img_raw

but the heading output will plot "-1.#.IO" when tracking uses the second camera, making post processing to crash.

Dario

NOTE: i don't think is related to this issue, but we are tracking using a timestamp we generate from the external video source.

@akramhussein
Copy link
Contributor

The 2nd camera looks to have a warped image. Was it correctly calibrated?
The first one looks fine.
For floor plan generation, the checkerboard must be placed in the centre of
an area that overlaps both cameras.

On 16 October 2013 05:03, drochairobot [email protected] wrote:

Hi Akram,

We have a floor plan that includes 2 cameras that have an approximate
orientation of 90 degrees between them. When creating the floor plan, it
shows a miss-alignment of the checkerboard:
[image: floorplanturning]https://f.cloud.github.com/assets/5649775/1339848/dfbe81e6-3615-11e3-8899-3569fdab0847.png

This will result on tracking miss aligned. The next image highlights a
trajectory that should be a straight line:
[image: track_result_img_out]https://f.cloud.github.com/assets/5649775/1339882/cda9d81a-3616-11e3-9603-dc7071fbd047.png

In the case we do not turn the image of the second camera when doing the
floor plan, the floor plan and trajectory will appear better aligned,
[image: track_result_img_raw]https://f.cloud.github.com/assets/5649775/1339901/b3a3a29c-3617-11e3-9ceb-ffe4c7b3f412.png

but the heading output will plot "-1.#.IO" when tracking uses the second
camera, making post processing to crash.

Dario

NOTE: i don't think is related to this issue, but we are tracking using a
timestamp we generate from the external video source.


Reply to this email directly or view it on GitHubhttps://github.com//issues/6
.

@drochairobot
Copy link
Author

The 2nd camera average reprojection error is 0.37. Only one of the 24 pictures used for calibration has a reprojection error of 0.8. The calibration was made according to procedure, but the data was imported to the new GTS release from the older version, since we don't have support for GigE cams.

The cameras are separated by a wall with a door in the middle, where the checkerboard is located and visible by both cameras.

@akramhussein
Copy link
Contributor

Can you show the unwarped images of each individual cameras calibration?
The previous post showed an unusual image for camera 2 behind the message
box. It may be worth sending a copy of the workbench without video files to
help track down the issue. If this is possible, please send them to akram
[dot] hussein [at] dyson [dot] com.

On 16 October 2013 22:13, drochairobot [email protected] wrote:

The 2nd camera average reprojection error is 0.37. Only one of the 24
pictures used for calibration has a reprojection error of 0.8. The
calibration was made according to procedure, but the data was imported to
the new GTS release from the older version, since we don't have support for
GigE cams.

The cameras are separated by a wall with a door in the middle, where the
checkerboard is located and visible by both cameras.


Reply to this email directly or view it on GitHubhttps://github.com//issues/6#issuecomment-26459164
.

@drochairobot
Copy link
Author

The unwarped image of the 2nd camera is:
unwarpedoutdoor

The unusual image you mention is always like that when the image is tilted to give the correct orientation of the camera, when doing the floor plan.

I might send the workbench to you soon.
Thanks!

@drochairobot
Copy link
Author

This is how it looks when creating the floor plan tilting the image:
creatingfloorplan

Without tilting it looks like this
creatingfloorplan2

@akramhussein
Copy link
Contributor

This image seems fine. Try retaking it with the checkerboard at 90 degrees. Then try to stitch them.

On 16 Oct 2013, at 23:40, drochairobot [email protected] wrote:

The unwarped image of the 2nd camera is:

The unusual image you mention is always like that when the image is tilted to give the correct orientation of the camera, when doing the floor plan.

I might send the workbench to you soon.
Thanks!


Reply to this email directly or view it on GitHub.

@akramhussein
Copy link
Contributor

Seems to be a bug with the rotation stage. If you rotate again to make it 180, does it look fine? Is this skewing only at 90 degrees?

On 16 Oct 2013, at 23:45, drochairobot [email protected] wrote:

This is how it looks when creating the floor plan tilting the image:

Without tilting it looks like this


Reply to this email directly or view it on GitHub.

@drochairobot
Copy link
Author

Exactly. 0 and 180 looks OK. 90 and 270 is tilted and looks split in the middle.

@akramhussein
Copy link
Contributor

Which version of software is this? Platform?

@drochairobot
Copy link
Author

0.4.4 on Windows.

@ghost ghost assigned akramhussein Oct 17, 2013
@akramhussein
Copy link
Contributor

Thanks. If I can get a copy of your workbench (please remove videos) that
would help me diagnose it. Is that possible?

On 17 October 2013 22:35, drochairobot [email protected] wrote:

0.4.4 on Windows.


Reply to this email directly or view it on GitHubhttps://github.com//issues/6#issuecomment-26555214
.

@akramhussein
Copy link
Contributor

I have tested 0.4.7 with the workbench you provided and it resolves this issue. Can you please confirm and then we can close this bug. Thanks

@drochairobot
Copy link
Author

Hi Akram, the alignment on the tracking looks OK now. Creating the room and the floor plan showed no issues.
There is still one issue remaining and that is that when tracking with the outside cameras with a different orientation, the heading plots the error value “-1.#I0”…

@drochairobot
Copy link
Author

It is doing this both when we turn and when we don’t turn the images at creating the floor plan for the room.

@User-Neha
Copy link

Hi,

Are you able to fix the issue of getting incorrect floor plan image for cameras in different room?
I have multiple cameras which are in different rooms, and i am getting absurd floor plan image using the latest version of the application. Can you please help by letting me know in case you did any specific change in the code for generating correct floor plan image.

Thanks in advance!
Neha

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

No branches or pull requests

3 participants