Extrinsic calibration

This a short tutorial for extrinsic calibration of Pepper's Depth and RGB cameras.

Setup

First of all, print a calibration pattern (a chessboard) if you do not have any. We would recommend A3 size. Once you have it, count the number of rows and columns and measure a size of a square. Adjust these parameters in launch/calibration_cameras_pepper.launch:

• define the numbers of columns -1
• define the numbers of rows -1
• define the size of one square in m

Data recording

• turn on your robot, wake up, and move the head to the "zero" pose, for example

  ssh nao@<PEPPER_IP>
  qicli call ALMotion.wakeUp
  qicli call ALMotion.setAngles "HeadPitch" 0.0 0.3

• take two post-it (one on the top of enougher) and fix it on the right eye of Pepper (to cover this eye with the camera inside)

• start Naoqi Driver, ensuring that color and IR images are enabled in share/boot_config.json

  roslaunch naoqi_driver naoqi_driver.launch nao_ip:=<PEPPER_IP>

• create an empty folder for data recording and provide its name in launch/extrinsic_calibration_pepper.launch

• update ROS topics for color and IR images in launch/extrinsic_calibration_pepper.launch

• launch the code:

  roslaunch extrinsic_calibration extrinsic_calibration_pepper.launch

• if the software can subscribe to ROS topics, then you should see an image from a color camera

• place the calibration pattern in front of the robot

• once the pattern is detected (you can see all detected crossings), press "s" on your keyboard to save images from both cameras

• repeat 2 previous steps each time changing a distance and/or an angle to cover different views (at 60cm-80m distance) until you reach a sufficient number of image pairs (for example, 30 times or more)

Calibration based on recorded data

• launch the code to compute extrinsic parameters based on recorded images:

  roslaunch extrinsic_calibration extrinsic_calibration_pepper.launch

• press "c" on your keyboard to compute calibration values

• check if the resulted XYZ and RPY for last processed image pair correspond to the reality

How to apply the extrinsic parameters

Once the calibration parameters are found, you can apply them using ROS static_transform_publisher. Check an example in register_depth.launch, update the values (x y z yaw pitch roll) and launch it after naoqi_driver:

roslaunch naoqi_driver naoqi_driver.launch nao_ip:=<PEPPER_IP>
roslaunch extrinsic_calibration register_depth.launch