The tag's coordinate frame is centered at the center of the tag, with x-axis to the right, y-axis down, and z-axis into the tag.
However, this contradicts the observations when running the apriltag_ros detector.
When I print a tag and look at it with a camera, the tag TF from the position of the camera is:
x towards the right of the camera
y upwards
z outwards towards the camera.
Likewise, when I create a tag bundle with several tags, the coordinate system used by the detector is:
x towards the right side of the paper
y upwards
This is very important when defining the position of each tag in the tag_bundles of the yaml file. Following a wrong convention leads to a completely wrong pose estimation.
Hi!
Following this issue, I would kindly recommend to clarify the convention used when defining the tag TF.
The only info I could find is here and says:
However, this contradicts the observations when running the
apriltag_ros
detector.When I print a tag and look at it with a camera, the tag TF from the position of the camera is:
Likewise, when I create a tag bundle with several tags, the coordinate system used by the detector is:
This is very important when defining the position of each tag in the
tag_bundles
of theyaml
file. Following a wrong convention leads to a completely wrong pose estimation.