Camera calibration by vision threads with pixel-resolved focus measurement
Final Report Abstract
Camera imaging systems are the sensory basis for many optical metrology techniques that find application in manufacturing quality control. Especially the measurement of 3D topography requires a quantitative geometric correspondence between real world and camera image. This correspondence is established by the so-called camera calibration, for which many methods and models exist. The generic camera model assigns to every camera pixel a vision ray, pointing in the direction of the light source. Therefore, arbitrary distortions can be described. Calibration of the generic camera model is a challenging task. In most cases, a reference target with a large number of markers is placed in different distances and orientations relative to the camera, such that every pixel observes multiple markers. By finding the positions of the reference target, the markers can be located in space and the desired vision rays can be fitted to them. Locating the reference target is a difficult numerical problem. The goal of this project is the development of a new calibration method for the generic camera model. The estimated marker points in space are connected to form so-called vision threads. The deviations of vision threads from straight lines are used to define forces and torques, which act in a computer simulation on the reference target’s estimated positions. This procedure is expected to improve the estimations of the reference target positions and consequently the calibration. We use a liquid crystal monitor as reference target and the markers are determined with phase shifting technique. The calibration method in a simulation environment was successfully implemented, even for the simultaneous calibration of multiple cameras. The method was however not feasible in the experiment and far inferior to the already existing Vision Ray Calibration method. An additional goal of this project was to use the fringe contrast that is gained during phase shifting to determine the focus distance for individual camera pixels. A mathematical model for the contrast degradation due to defocus was derived, which enables the determination of the focus distances experimentally with an estimated accuracy in the order of millimeters, or approximately 1% of the focus distance.
Publications
-
Sichtfäden - Ein neuer Ansatz zur generischen Kamerakalibrierung. Proc. DGaO, Bd. 122, B007, 2021
Shishir Gauchan, Jonas Bartsch, Ralf Bernhard Bergmann
-
Vision threads: a novel approach to generic camera calibration. Automated Visual Inspection and Machine Vision IV (2021, 6, 20), 15. American Geophysical Union (AGU).
Gauchan, Shishir; Bartsch, Jonas & Bergmann, Ralf B.
-
Extending vision ray calibration by determination of focus distances. Optics Express, 30(26), 47801.
Sperling, Yann; Bartsch, Jonas; Gauchan, Shishir & Bergmann, Ralf B.
