Automatic Selection for Optimal Calibration Model of Camera

When calibration model express camera status well, calibration result will be accurate. How to select optimal calibration model for different camera by program is significant to calibrate automatically. A method of selecting optimal calibration model was proposed in this paper. First, the models including enough possible status were selected from physical models and Chebyshev models. These models would be taken as the input of our method. Second, candidate calibration models were obtained by variance detection. Third, optimal calibration model was extracted by utilizing detection of minimum description length. Experimental results show that calibration residuals is lowest via optimal model, computation error of principle distance is no more than 0.2 pixels and variance lies in the bounds of 0.5 pixels.