Distortion model selecting and accuracy evaluation for CCD camera calibration

The precision of a CCD camera calibration depends on both the distortion models selected and the signal-to-noise. The former is called the intrinsic factor, and latter is the extrinsic factor. So far, none of the fixed criterion has been used for selecting the distortion model, and almost all are personal or experiential. None of the fixed criterion has been used for determining the degree of error with which we measure the 2D image coordinates, and the amount by which the calibration model can correctly compensate for the distortion in a fixed imagery system in current calibration methods. The paper presents notable and correlation coefficient checks in the statistics for selecting the distortion model, using the signal-to-noise ratio as the criterion to measure the compensation capability. A large number of simulation and practical experiments show: (1) various distortion models exist for imagery systems; (2) there is a high correlation among the distortion parameters or between the distortion parameters and the camera parameters; (3) the larger the signal-to-noise ratio, the larger the compensation capability, and the smaller signal-to-noise ratio, the smaller the compensation capability; and (4) the relation between the precision improvement and the signal-to-noise ratio is an exponential function.