KinectV2 Python Calibration

Easy and fast camera KinectV2 camera calibration using Python 3.

This project is an updated fork of @tataraidze KinectV2 Camera calibration scripts. He describe his work, in Russian, in this link here (in Russian). An English translation can be found here.

Dependencies

1. OpenCV 3.0 with Python 3 support
2. PyKinect2

Installing on Windows

I recommend using Anaconda's 32 bits Python3 distribution. It bundles packages such as comtypes and numpy.

1. Install Python 3 32bits from Anaconda's download page

   Make sure to register Anaconda as your default path. pykinect2 will not work on the 64bits distro (more info here).

2. Clone and install pykinect2

   1. pip install comtypes
   2. git clone https://github.com/Kinect/PyKinect2.git
   3. python -m easy_install PyKinect2

3. Install OpenCV 3.3.1 (or newer from https://www.lfd.uci.edu/~gohlke/pythonlibs/).

   1. Download the right .egg file for your python version (e.g.: 3.6) and architecture (e.g.: 32bits). In my case I downloaded, opencv_python‑3.3.1+contrib‑cp36‑cp36m‑win32.whl
   2. Install it with pip

   pip install opencv_python-3.3.1+contrib-cp36-cp36m-win32.whl

File descriptions

• const.py - constants: paths, pattern size, e.t.c.
• takePictures.py - calibration data collecting
• calibrate.py - single camera calibration (RGB or IR/depth)
• testDistortion.py - check whether distortion was removed or not
• stereoCalibrate.py - stereo camera calibration finds transformation |R,T| between depth camera space and RGB camera space
• depthCalibration.py - depth sensor calibration
• KinectV2.py - wrapper to PyKinect2 compensates distorion and systematic error of depth sensor; combines depth, RGB and bodyIndex frames; saves data
• sample.py - small example of using KinectV2.py

How to use

1. Change constants in const.py
2. Collect data for geometric calibration by takePictures.py (20-30 images of object in different positions)
3. Calibrate RGB camera by calibrate.py
4. Calibrate IR/depth camera by calibrate.py
5. Find |R,T| transformation between depth camera space and RGB camera space by stereoCalibrate.py
6. Collect data for depth calibration by takePictures.py ()
7. Use sample.py to get corrected data

Support for this work was provided by the Russian Science Foundation, project #15-19-30012. You can find more about the project [here] (http://www.rslab.ru/russian/project/rsf2/)