Improving the Generic Camera Calibration technique by an extended model of calibration display

Generic camera calibration is a method to characterize vision sensors by describing a line of sight for every single pixel. This procedure frees the calibration process from the restriction to pinhole-like optics that arises in the common photogrammetric camera models. Generic camera calibration also enables the calibration of high-frequency distortions, which is beneficial for high-precision measurement systems. The calibration process is as follows: To collect sufficient data for calculating a line of sight for each pixel, active grids are used as calibration reference rather than static markers such as corners of chessboard patterns. A common implementation of active grids are sinusoidal fringes presented on a flat TFT display. So far, the displays have always been treated as ideally flat. In this work we propose new and more sophisticated models to account for additional properties of the active grid display: The refraction of light in the glass cover is taken into account as well as a possible deviation of the top surface from absolute flatness. To examine the effectiveness of the new models, an example fringe projection measurement system is characterized with the resulting calibration methods and with the original generic camera calibration. Evaluating measurements using the different calibration methods shows that the extended display model substantially improves the uncertainty of the measurement system.