For photogrammetric works, a fundamental issue is the determination of camera internal orientation parameters (IOP). Without camera calibration, it is difficult to imagine a correct adjustment of the image network. Many industries use nonmetric cameras, ranging from automatics and robotics, to heritage inventories, and the increasingly popular social mapping phenomenon uses low-budget cameras. Many different calibration methods exist, but dedicated calibration fields are commonly replaced by fast in-plane calibration with regular patterns. The main goal of this research is to verify the thesis that calibrating cameras on a checkerboard gives worse results in determining IOP than on a laboratory test field which may translate into the resulting model. For the purpose of this study, a special field was constructed, allowing calibration of the instruments on the basis of the network solution by the bundle adjustment. Unlike classical 2D fields, the field is equipped with a cork background providing a good base for matching and automatically detecting measurement marks. Calibration results were compared with calibration performed on a checkerboard implemented in MATLAB Camera Calibration Toolbox. In order to determine IOP in MATLAB, images of the checkerboard must be taken in such a way, that the whole pattern fits into the frame, otherwise toolbox defines the incorrectly coordinate system, which has a bad impact on calibration results. Moreover, the determined parameters have several times larger standard deviations than those determined in the laboratory test field, which confirms the thesis.