aruco_board_detect

ROS package to detect aruco boards and markers in images, estimating their pose. Written in C++ and Python.

Install tips

This package depends on OpenCV and the ArUCO library bundled with it. You might want to build OpenCV and opencv-contrib from source, since fiducial marker detection is somewhat dodgy in OpenCV<3.2, and OpenCV 3.2 is bundled with Ubuntu Bionic and ROS Melodic.

export OpenCV_DIR=your/opencv/build/or/install/path
catkin build -DOpenCV_DIR=${OpenCV_DIR}

or

export OpenCV_DIR=your/opencv/build/or/install/path
catkin_make -DOpenCV_DIR=${OpenCV_DIR}

to build the workspace. On Ubuntu Bionic I had to compile from source the dependencies of this package that also rely on OpenCV:

common_msgs (jade-devel)
image_common (hydro-devel)
vision_opencv (melodic)

Usage

Once a camera is plugged in your rig and its node is running, simply use the roslaunch file provided. Some roslaunch parameters you might find useful:

Parameter	Effect
camera_info_topic	The node will subscribe to this topic to source camera parameters
camera_image_topic	The node will subscribe to this topic to source input images
show_debug_img	Shows the output image in a window
publish_single_markers	Enable single marker detection
detection_rate	Time (second) between marker detections
board_config_file	Config file for the marker board
single_markers_config_file	Config file for the single markers

You can change the configuration of the marker board being sought in the cfg/board_config.yaml file. Same goes for individual markers.

The package also contains a script to generate fiducial markers.

Expected output

The node will broadcast a tf frame named aruco_board (the board reference frame) and one named graspa_board (the same frame, shifted on the bottom right of the board).

Publish

Topic	Explanation
/aruco_board_detector/board_pose	The marker board pose (stamped with the camera reference frame)
/aruco_board_detector/debug_image	The output image, i.e. the input image with markers drawn on it
/aruco_board_detector/marker_data	Stamped 6D pose of every single marker detected, with ID

Subscribe

Topic	Explanation
camera_info_topic	Camera parameters topic
camera_image_topic	Camera image topic

Example command

roscore
roslaunch realsense2_camera rs_rgbd.launch
roslaunch aruco_board_detect aruco_board_detect.launch publish_single_markers:=true show_debug_img:=true

Script example command

The following command can be used to generate a 400x400 px marker (ID 44) with a 100 px white border (quiet zone) and a 50 px black border. DICT_4X4_50 is the dictionary, as defined by the OpenCV headers.

python `rospack find aruco_board_detect`/scripts/generate_aruco.py -o marker_44.png -w 100 -b 50 -s 400 -i 44 -t DICT_4X4_50

To use for Kinova arm

1. Clone from https://github.com/Kinovarobotics/ros_kortex_vision, https://github.com/KishanSawant/aruco_board_detect and https://github.com/Kinovarobotics/ros_kortex
2. Set ROS_IP on local system to 192.168.1.11
3. In CMakeLists.txt of aruco_board_detect, make this change,

   find_package(OpenCV 4.2 REQUIRED)
   

4. In aruco_board_detect.launch, make this change,

   <arg name="show_debug_img"      default="true" />
   

5. After exporting ROS_IP, launch,

   roslaunch kortex_driver kortex_driver.launch device:=192.168.1.12
   roslaunch kinova_vision kinova_vision_rgbd.launch device:=192.168.1.12
   

6. In another terminal, launch,

   roslaunch aruco_board_detect aruco_board_detect.launch
   

7. Save the data from:

   rostopic echo /aruco_board_detector/board_pose
   rosrun tf tf_echo /base_link /camera_link
   

8. Process this data to estimate the global frame with respect to base_link of respective arm
9. PS: use aruco board instead of single marker