No tf camera frame neither camera_info are published about pylon-ros-camera

[BruttherJOE]

Hi @bach05, I am currently facing the same issue with 2 basler ace cameras and on Noetic as well. Let us update on this thread if we have any breakthroughs

[bach05]

My hypothesis it's we need a common parent frame (e.g. a fixed world frame) for both cameras. Otherwise tf is not able to manage both cameras. In fact if you run onyl ona camera, the tf will be published.

I am going to calibrate extrinsics of both cameras wrt a fixed point to validate this hypothesis.

[BruttherJOE]

I used the ros stereo cameracalibrator.py to calibrate the intrinsics and extrinsics.

How are you able to get tf with only one camera? did you do monocular calibration?

My rqt graph looks like this :

and I am receiving images on rect too. I republished camera_info for stereo/left and stereo/right from another topic that provides it, /stereo/leica_basler_left/pylon_camera_node/camera_info and that solves the problem of camera_info but image frame does not match?

I am new to stereo vision and I am quite possibly doing everything wrongly

[bach05]

How are you able to get tf with only one camera? did you do monocular calibration?

I used the ros stereo cameracalibrator.py to calibrate the intrinsics and extrinsics.

How are you able to get tf with only one camera? did you do monocular calibration?

Yes, you can do monocular calibration with a single camera.

and I am receiving images on rect too. I republished camera_info for stereo/left and stereo/right from another topic that provides it, /stereo/leica_basler_left/pylon_camera_node/camera_info and that solves the problem of camera_info but image frame does not match?

It's a bit hard to debug. There is a mismatch between the camera frame (in camera info header) and the image frame (in the image topic). Probably you should check the nodes that are publishing those information to fix the message headers in order to have the same frame. If you need to visualize the images only, you can just use the Image tool on Rviz.

[BruttherJOE]

Ok, I will try to fix the message headers. The end goal for me would be to get a pointcloud2 out on Rviz with the stereo camera setup. By the way, I have also tried to link back to /world like you have mentioned, but I think I need to fix my headers first.

[FrancoisPicardDTI]

Hello @bach05 and @BruttherJOE
Thank you for looking at this issue. Did you manage to implement some fixes at the end?
Best regards,
Francois

[BruttherJOE]

hello @bach05 @FrancoisPicardDTI,

I believe I did manage to fix my headers, and I was able to get camera_info and tf_frames from my calibrated files. However, the rectified images looked a bit wrong and was zoomed in wrongly possibly due to my calibration (I am using some custom stereo setup which goes through multiple lenses). Also, I was not able to get a pointcloud, perhaps due to this reason. Perhaps I do not know enough to debug these problems, but I calibrated my cameras in another way with matlab in the end as we can choose other (objectively better?) calibration patterns for our needs and converting / importing the parameters into opencv. The idea of using ros still interests me, so I will take a look at this in my free time.

Best Regards,
Joe