Calibration toolkit

A. Fusiello

Description

Camera calibration determines the external and internal parameters (including radial distortion coefficients) of a camera given a certain number of images of a calibration object, whose dimensions are known. Two techniques are available:

• Resection:  classical technique [1], uses a single picture of a non-planar reference object;
• SMZ: the Sturm-Maybank-Zhang [2,3] calibration algorithm, that uses multiple pictures of a planar object (a checkerboard). This is the same algorithm implemented in OpenCV and in the "Camera Calibration Toolbox for Matlab" by Jean-Yves Bouguet.

Reference points are detected with template matching on a (perspective) rectified image; four points for each plane must be specified by the user in a predefined order by clicking on the image. Several parameters can be changed by editing the two main scripts: documentation should be clear enough to allow for customization.

The last version includes also an automatic procedure based on AprilTags detection (it requires the apritag Python module).

Code

The MATLAB Calibration Toolkit is based on functions contained in the MATLAB  Computer Vision Toolkit by A. Fusiello. The code does not have any other external dependency on toolboxes or packages, and runs indifferently on Octave and Matlab. It has been tested on Octave 4.2.1 and Matlab R2017a. The checkerboard can be downloaded from here

References

1. Richard Hartley and Andrew Zisserman. 2003. Multiple View Geometry in Computer Vision (2nd. ed.). Cambridge University Press, USA.
2. Zhengyou Zhang. 2000. A Flexible New Technique for Camera Calibration. IEEE Trans. Pattern Anal. Mach. Intell. 22, 11 (November 2000), 1330 - 1334.
3. Sturm, Peter F. and Stephen J. Maybank. On plane-based camera calibration: A general algorithm, singularities, applications. Proceedings. IEEE Comp. Soc. Conf. on Computer Vision and Pattern Recognition (1999): 432-437