🎯 USB Camera Intrinsic Calibration (Camera Calibration)

1. Course Content

Obtain the camera intrinsic parameters of the USB camera through checkerboard calibration board, used for publishing camera camera_info messages.

2. Preparation

2.1 Prepare Checkerboard Calibration Board

• Prepare a checkerboard calibration board, download and print the corresponding camera calibration board from the supporting materials [Attachments - Camera Checkerboard Calibration Board] and attach it to cardboard or other flat surface

2.2 Start Camera Driver and Calibration Program

Open the terminal and run the camera intrinsic calibration launch file:

ros2 launch dofbot_pro_driver A1pro_camera_arm.launch.py

Run the calibration node (specify internal corner points and physical square size):

xxxxxxxxxx
ros2 run camera_calibration cameracalibrator --size 9x6 --square 0.024 --ros-args -r image:=/usb_cam/image_raw

3. Calibration Operation Flow

3.1 Calibration Interface Description

The right side of the calibration window shows four progress bars:

• X: Sample coverage for horizontal movement (left/right) of the checkerboard in the camera's field of view

• Y: Sample coverage for vertical movement (up/down) of the checkerboard in the camera's field of view

• Size: Coverage for depth distance as the checkerboard moves closer/farther from the camera

• Skew: Coverage for tilt, pitch, and roll rotation angles of the checkerboard in the field of view

3.2 Sampling and Calculation Steps

1. Hold the calibration board in front of the camera's field of view, slowly adjust the board position (up/down, left/right, closer/farther, and tilt/pitch/rotation);

2. When the algorithm successfully captures all internal corners, colored corner connection lines will be drawn in the image, and the corresponding progress bars on the right will change from red to yellow, and finally to green;

3. When all progress bars turn green and the CALIBRATE button in the lower right corner of the interface is lit and activated, click CALIBRATE to execute intrinsic parameter optimization (this process takes tens of seconds, please be patient);

4. After the solution is complete, the terminal will print the camera intrinsic matrix and distortion error. Click SAVE to save the calibration data to /tmp/calibrationdata.tar.gz;

5. Click COMMIT to submit and exit.

   

4. Export and View Calibration Results

4.1 Extract Calibration Package

Execute the extraction command in the terminal:

xxxxxxxxxx
mkdir -p ~/camera_calibration_data
tar -zxvf /tmp/calibrationdata.tar.gz -C ~/camera_calibration_data/
cd ~/camera_calibration_data/
ls -la

After extraction, the directory contains all collected sample images, ost.txt, and the standard intrinsic file ost.yaml, which is the camera's intrinsic parameters.

4.2 Fill the Calibrated Intrinsics into These Two Files

• Camera Startup Parameters

  Startup command,

  ​x
  cd ~/camera_calibration_data/
  cp ost.yaml ~/yahboomcar_ros2_ws/software/library_ws/src/usb_cam/config/camera_info.yaml
  
  # After replacement, compile
  cd ~/yahboomcar_ros2_ws/software/library_ws
  colcon build --packages-select usb_cam
  

• Spatial 3D Grasping Camera Parameters

  It is recommended to modify through vscode remote login, modify the following according to the intrinsic parameters from ost.yaml above:

  xxxxxxxxxx
  gedit  ~/yahboomcar_ros2_ws/dofbot_pro_ws/src/dofbot_pro_usb/config/usb_camera_params.yaml
  

  Modify the parameters above

  

5. Camera Intrinsic YAML Structure Analysis

Standard Intrinsic File Reference:

xxxxxxxxxx
~/yahboomcar_ros2_ws/software/library_ws/src/usb_cam/config/camera_info.yaml

xxxxxxxxxx
image_width: 640
image_height: 480
camera_name: narrow_stereo
camera_matrix:
  rows: 3
  cols: 3
  data: [567.38936,   0.     , 307.17636,
           0.     , 568.87412, 227.58214,
           0.     ,   0.     ,   1.     ]
distortion_model: plumb_bob
distortion_coefficients:
  rows: 1
  cols: 5
  data: [-0.435109, 0.246398, 0.002171, 0.006236, 0.000000]
rectification_matrix:
  rows: 3
  cols: 3
  data: [1., 0., 0.,
         0., 1., 0.,
         0., 0., 1.]
projection_matrix:
  rows: 3
  cols: 4
  data: [487.55125,   0.     , 309.493  ,   0.     ,
           0.     , 523.12151, 226.38238,   0.     ,
           0.     ,   0.     ,   1.     ,   0.     ]

• image_width / image_height: Camera image resolution, here 640×480 pixels

• camera_name: Camera name identifier

• camera_matrix: 3×3 camera intrinsic matrix K = [fx, 0, cx; 0, fy, cy; 0, 0, 1]

  • fx, fy: x/y direction focal length (pixel units), determines image scaling

  • cx, cy: Principal point coordinates (pixel position of optical center in image), ideally close to image center

• distortion_model: Distortion model, plumb_bob is a 5-parameter distortion model (k1, k2, p1, p2, k3)

• distortion_coefficients: Distortion coefficients

  • k1, k2, k3: Radial distortion coefficients, describing barrel/pincushion distortion caused by lens curvature

  • p1, p2: Tangential distortion coefficients, describing distortion caused by lens installation deviation

• rectification_matrix: 3×3 rectification matrix, for monocular camera it is identity matrix (no stereo rectification needed)

• projection_matrix: 3×4 projection matrix P = [fx', 0, cx', Tx; 0, fy', cy', Ty; 0, 0, 1, 0]

  • For monocular camera, Tx=0, Ty=0, used to project 3D points to 2D image plane