Camera calibration

We use a simple checkerboard to obtain camera intrinsic properties and distortion parameters.

Run

Create virtual environment

It's advised to create a conda environment first.

conda create -n camera python=3.8
conda init
exec zsh
conda activate camera

Clone repository

git clone https://github.com/yashdeep01/CameraCalibrationCheckerboard.git
cd CameraCalibrationCheckerboard/

Install requirements

pip install -r requirements.txt

File structure

Cloning the repo must give you such a directory tree

├── CameraCalibrationCheckerboard
│   ├── README.md
│   ├── calibrate.py
│   ├── checkerboard/
│   ├── extract_frames.py
│   ├── frames/
│   └── requirements.txt

Populate directories

Before running calibration on your machine, you must click some pictures of the chekcerboard given as pattern.pdf here. Follow the steps:

1. Print the checkerboard pattern on an A4 sheet.
2. Glue the pattern on a flat surface, like a wall, cardboard, etc.
3. Keep the surface still and click its photos from various angles. Click at least 50 images. Alternatively, you could also record a video by moving the camera around the surface at various angles and distance.
4. Save the photos (or video) on your machine.
   • If you clicked pictures, save the clicked pictures in ./frames/ directory of the cloned repo.
   • Else, if you recorded a video, run this command to extract frames from the video and automatically save in ./frames/ directory:

   python extract_frames.py --video /path/to/video

Run camera calibration

python calibrate.py

Aruco markers drawn on the checkerboard photos are stored in ./checkerboard/ directory.

The results of the calibration command are present in ./results/ directory.

• ./results/mtx.csv contains the 3x3 camera intrinsic matrix (denoted by K).
• ./results/dist.csv contains the distortion parameters k1, k2, k3... which are responsible for the fish-eye effect in the images captured by the camera.
• ./results/rvecs.csv contains the rotation vectors R of the camera extrinsic matrix.
• ./results/tvecs.csv contains the translation vectors t of the camera extrinsic matrix.

Optional commands

This package allows users many flexibilities while performing calibration. Running python calibrate.py --help gives the following output on shell:

 % python calibrate.py --help
usage: calibrate.py [-h] [-p PATH] [-r ROW] [-c COL]

Calibrate a camera using checkerboard

optional arguments:
  -h, --help            show this help message and exit
  -p PATH, --path PATH  Path of the images directory
  -r ROW, --row ROW     No. of rows of the checkerboard
  -c COL, --col COL     No. of cols of the checkerboard

Acknowledgements

• OpenCV camera calibration docs.
• Checkerboard pattern provided by boofcv.