camera_hand_eye_calibration

ROS node for pinhole camera calibration and hand eye calibration.

Features

Compared with other camera calibration tool (e.g., camera_calibration)and hand-eye calibration tool (e.g., easy_handeye), this package provides the following features:

• Camera and hand-eye position calibrated together. The user only need to move the camera (by commanding the robot) around the target and collect a single set of data. The data will be used for both camera calibration and hand-eye calibration.
• Use ArUco boards instead of the traditional chessboards. Detecting of AruCo board are much more versatile because it allows occlusions and partial views.

Install Latest OpenCV

The custom built OpenCV that comes with ROS melodic will not work properly for detecting AruCo board. We need to build latest OpenCV from source.

sudo apt update && sudo apt install -y cmake g++ wget unzip
# create a folder for OpenCV
mkdir ~/OpenCV && cd ~/OpenCV
# Download and unpack sources
wget -O opencv.zip https://github.com/opencv/opencv/archive/master.zip
wget -O opencv_contrib.zip https://github.com/opencv/opencv_contrib/archive/master.zip
unzip opencv.zip
unzip opencv_contrib.zip
# Create build directory and switch into it
mkdir -p build && cd build
# Configure
cmake -DOPENCV_EXTRA_MODULES_PATH=../opencv_contrib-master/modules ../opencv-master
# Build
cmake --build .

It might take 20 minutes to build. See here if needs more instructions. At the time when I created this repository, the lastest OpenCV version is 4.5.1

Usage

1. prepare the ChArUco Board

2. create launch file

<launch>
    <!-- start your robot driver -->

    <!-- start your camera driver -->

    <!-- start calibration node -->
    <!-- you need to replace the argument value based on your application -->
    <include file="$(find camera_hand_eye_calibration)/launch/launch_calibration.launch">
        <arg name="image_topic"        value="/avt_camera_img"   doc="the image topic name"/>
        <arg name="tf_gripper_name"    value="tool0_controller"  doc="name of the gripper joint in the tf tree"/>
        <arg name="tf_base_name"       value="base"              doc="name of the robot base in the tf tree"/>
        <arg name="chessboard_rows"    value="10"                doc="number of chess board rows"/>
        <arg name="chessboard_columns" value="8"                 doc="number of chessboard columns"/>
        <arg name="square_length"      value="0.02244"           doc="chessboard square length (meter)"/>
        <arg name="marker_length"      value="0.01122"           doc="Aruco marker length (meter)"/>
    </include>
</launch>

3. collect data

4. calibrate

The calibration results will be print to the console, also saved to yaml file "camera_hand_eye_calibration.yaml". You can locate the file in ~/.ros folder.

# This is autogenerated file to store camera and hand-eye calibration results
camera_calibration:
  # camera matrix
  K: [1350.881910478003, 0, 768.9209200452633, 0, 1350.994460435373, 556.015338127424, 0, 0, 1]
  # distortion parameters
  D: [-0.07149669652986015, 0.1038887256802538, -0.0007713606820135701, 0.0009113898706119508, -0.05958660621785733]
hand_eye_position:
  # rotaion matrix
  rotation: [-0.000466712, -0.00562231, -0.00715545, 0.999958]
  # translation
  translation: [-0.001117274244385104, -0.08463651745339704, 0.09354036398990093]

the hand eye position is the transformation of camera frame with respect to end-effector frame.