No computed baseline value set in /camera_info topic in the recorded rosbag file for right infrared sensor

[vanguard478]

---

Required Info
Camera Model	D400 (D435i)
Firmware Version	(Open RealSense Viewer --> Click info)
Operating System & Version	Ubuntu 18.04
Kernel Version (Linux Only)	5.4.0-121
Platform	PC
librealsense	2.51.1 RELEASE
Language	C++
Segment	Others

Issue Description

Hi, I am trying to record a ros bag file of both Left and Right Infrared images using the rs-record utility. So to enable the infrared streams I modified the rs2::config cfg as follows to suit my needs:

   cfg.enable_stream(RS2_STREAM_INFRARED, 1,848,480, RS2_FORMAT_Y8, 15);
   cfg.enable_stream(RS2_STREAM_INFRARED,2,848,480, RS2_FORMAT_Y8, 15);
   cfg.enable_record_to_file(out_file.getValue());

As mentioned in the following issue in the realsense-ros repo, the Projection matrix should have the P[0,3] = -fx * B in the Camera Info topic (according to sensor_msgs/CameraInfo ) I recorded the bag file using the modified rs-record and there is no corresponding value for the computed -baseline*fx value in the Right Infrared Sensor Camera Info topic. The values are same for both left and right infrared camera topics. I even cross-checked with the realsense-viewer to record a rosbag file, and it also has the same issue for the Camera Info topic. Although, the baseline value can be queried in the ´/device_0/sensor_0/option/Stereo_Baseline/value´ topic, it makes the bag file recorded by rs-record difficult to use with some SLAM algorithms which depend on the Camera_Info topic to query the baseline.

Is there any parameter in the rs::config I can change so that the camera_info topic value will be correctly set in stereo recording setup?

[MartyG-RealSense]

Hi @vanguard478 I believe that this issue has been discussed on the RealSense ROS forum over several different cases - including https://github.com/IntelRealSense/realsense-ros/issues/1310 that you quoted - and so it can be a difficult subject to follow due to one case leading to another and then another.

It may therefore be best to link to a pull request at https://github.com/IntelRealSense/realsense-ros/pull/1242 that was eventually created and merged into the RealSense ROS wrapper at the end of the lengthy discussions.

The comments in this discussion remark that infra2 used the frame_id of infra1. This is because having the right camera frame_id equal to that of the left camera is required by StereoCameraModel from image_geometry and stereo_image_proc

This subject has been further discussed at https://github.com/IntelRealSense/realsense-ros/issues/2220#issuecomment-1010720769 where suggestions for dealing with it were offered.

The information applies to the RealSense ROS wrapper though. I am not aware of a method by which the issue could be addressed through librealsense SDK scripting, unfortunately.

[MartyG-RealSense]

Hi @vanguard478 Do you require further assistance with this case, please? Thanks!

[vanguard478]

Hi, I referred to the issues which you pointed out and I think Camera Info matrix are properly set by the ROS wrapper. The issue with rs-record is still there. As of now I am just modifying the published message as per my requirement.

[MartyG-RealSense]

Thanks very much @vanguard478 for the update

[MartyG-RealSense]

Hi @vanguard478 Can this case be closed now or do you have further questions, please? Thanks!

[vanguard478]

Hi @MartyG-RealSense, I don't have any further question regarding this. I just hope this issue with the Camera Parameters being incorrectly set in the bag file recorded by the rs-record utility is fixed in an future release.