search

pal-robotics / aruco_ros

Software package and ROS wrappers of the Aruco Augmented Reality marker detector library
MIT License
439 stars 304 forks source link

aruco_ros on noetic only detects AprilTag and original ArUco markers #110

Closed joaopsousa99 closed 1 year ago

joaopsousa99 commented 1 year ago

I'm running aruco_ros with the command roslaunch aruco_ros single.launch markerId:=0 and testing the detection with a marker from several different dictionaries available in OpenCV (Python 3's cv2), all with ID 0. The dictionaries I am using are:

All the AprilTag and the ArUco Original dictionaries are being detected just fine. The others are not detected at all.

After some research, I thought the problem was due to me having generated all the cv2.aruco.DICT[4:7]X[4:7] with the _50 at the end, which, according to this, corresponds to a minimum hamming distance of 4, 8, 13 and 19, respectively. I tried using cv2.aruco.DICT_4X4_250, cv2.aruco.DICT_5X5_100 and cv2.aruco.DICT_6X6_250 instead (no 7X7 option available for aruco_ros), which, according to this, correspond to aruco_ros's ARUCO_MIP_16h3, ARUCO_MIP_25h7 and ARUCO_MIP_36h12.

However, when I looked at the resulting markers, they are equal to the ones I had already generated (the ones from the dictionaries ending in _50).

This leads me to think the problem lies on the package and not the markers.

OS: Ubuntu 20.04 LTS ROS: Noetic

saikishor commented 1 year ago

Hello @joaopsousa99,

What's the problem exactly?. As per the title, the package does detect ArUco markers and we don't see the issue here. Please try to post the info so we can reproduce it on our end.

Thank you,

Best Regards, PAL Robotics Team.

joaopsousa99 commented 1 year ago

Sorry, I'll try to be clearer.

The issue is that, out of all the dictionaries I listed, the only only detects:

To reproduce the error:

  1. Start camera drivers
  2. roslaunch aruco_ros single.launch markerId:=0
  3. Point camera at a marker with ID 0 from either the 4X4, the 5X5, the 6X6 or the 7X7 dictionaries
  4. Point camera at a marker with ID 0from either the DICT_ORIGINAL or one of the AprilTag dictionaries
  5. See that in step 4 nothing is detected
  6. See that in step 5 any marker is correctly detected
saikishor commented 1 year ago

Hello @joaopsousa99,

This library only works with the Original ArUco markers. It doesn't support other marker types. They are completely different information encoded into them. You can generate your tag using : https://chev.me/arucogen/ choosing Original ArUco in the drop down. Moreover, conceptually it doesn't make sense for us to support different marker dictionaries, imagine you are interested in getting the information of Marker 0, and then you find 3-4 markers of different types in the scene, you don't want the pose estimation to jump from one marker to another.

Thank you,

Best Regards, PAL Robotics Team.

joaopsousa99 commented 1 year ago

I see, I thought the package worked with all ArUco dictionaries available in OpenCV. Thank you for you help.

As a sidenote: why does the package support AprilTag, then?

saikishor commented 1 year ago

Hello @joaopsousa99,

As a sidenote: why does the package support AprilTag, then?

Regarding this we are unsure, we can dig into this in future. Thanks for bringing it to our notice.

Best Regards, PAL Robotics Team.