Closed ibaranov-cp closed 5 years ago
T265 has different coordinate system then the D400 series. You can compare D400: T265:
Yep, understood. However the camera frames for left and right should still not be identical.
That is correct. The extrinsics are not available yet by the device. The TF values will be fixed automatically, as soon as the firmware is fixed.
This had been fixed. What can be seen here are the two fisheye optical frames together with the T265 body frame (in the center). Thanks for filing the issue, @ibaranov-cp!
Great, thanks!
@doronhi @schmidtp1 What confuses me is that if we treat the camera_pose_frame diagram in rviz the same as the one in the image, the camera's align on the y axis in rviz while in the image they align on the x axis. Could you please provide a diagram with the camera and the imu_frame fisheye frame and body frame similar to the one below from vins-fusion:
@doronhi @schmidtp1 What confuses me is that if we treat the camera_pose_frame diagram in rviz the same as the one in the image, the camera's align on the y axis in rviz while in the image they align on the x axis. Could you please provide a diagram with the camera and the imu_frame fisheye frame and body frame similar to the one below from vins-fusion:
sorry, i just saw this again now... i think the confusion comes from the different reference/body frames used in ROS/rviz and T265/librealsense. T265 uses a coordinate system as depicted in the first diagram ("VR coordinates") with y pointing up, z aligned with the optical axis and x with the baseline of the cameras, that is exposed by librealsense. Here in the ROS wrapper there is a transformation that rotates the frame to follow ROS/robotics convention with z pointing up etc. as can be seen in the screenshot from rviz. hope this helps.
TF values for T265 are incorrect, they have the left and right frames exactly co-located, and camera link to optical frame is backwards.