This report concerns the calibration of a visual-light camera. In order to calibrate the camera, a pin-hole camera model is used. This model can be used for conversion between real coordinates and image coordinates, and between image coordinates and sight lines in real space. The calibration is needed for a camera which is used on TCV, specifically in the OFIT algorithm.
The purpose of this research is to find an optimization process for the fourteen camera parameters of the pin-hole camera model. This is realized in two steps. First, the eight intrinsic camera parameters are optimized using a picture of a sheet of paper taken from an approximately known distance. On this paper, various lines and concentric circles are printed, all with a known distance from the center of the paper. Using the known real coordinates of the various intersection points between the lines and circles, and the known image coordinates of these points, the intrinsic parameters are successfully optimized.
The second step is to make a picture with the camera in its final position. In this case, the final position of the camera is on one of the tangential ports of TCV. Since the intrinsic parameters are known, now solely the extrinsic parameters are optimized. Using a picture of the interior of TCV and the known real-space coordinates of various points in the interior of the tokamak, also the extrinsic parameters are successfully optimized.
In both situations, the real coordinates are transformed to image coordinates using an initial guess of parameters. The computed image coordinates are compared to the real image coordinates and the difference gives a quadratic error. By automatically changing the camera parameters, this error decreases until a (sometimes local) minimum is reached, giving the best set of parameters. Since the error minimum can be local, a good initial guess is necessary.
A last step in this project is to implement this method for OFIT purposes. The optimization process described above is successfully executed. An additional step is, however, necessary for practical implementation. The resolution of the images and the coordinate system has changed from calibration to implementation, requiring a transformation from the optimized set of parameters to the new system. The optimized set of parameters fit the real time process pretty well, concluding that this transformation is also successful.
A last remark is that some parameters might change during operation on TCV or any other tokamak, due to the effect of magnetic fields. This causes change in mainly the angles and sometimes the optical center. In order to compensate this effect either dynamic calibration or a better magnetic shielding can be used. That is, however, not part of this project.
Minor project. - CST 2013.029