Research on the Influence of Calibration Image on Reprojection Error

This paper aims to capture calibration pictures by adopting three shooting methods at arbitrary angles, micro angles, and fixed angles, conducting camera calibration tests on data from multiple sets of pictures, and determining the average reprojection error of every dataset. Therefore, it is possible to evaluate the effects of calibration boards, number of calibrated pictures, and shooting methods of the calibrated pictures on calibration results. Based on the change curve of reprojection errors, it can be concluded that calibration effects of the circular array calibration board are more stable than those of the checkerboard; however, the impact is not significant. If there are a few calibration pictures, camera parameters will evidently be prone to a certain error rate. Moreover, too many calibration pictures will not only escalate the computation load but also result in a certain accumulation of errors and multiply the probability of error. The shooting method of a calibration image can greatly affect the calibration outcome. In other words, the standardization of a shooting image plays a central role in reducing reprojection errors and improving the calibration accuracy.