A novel circular points-based self-calibration method for a camera's intrinsic parameters using RANSAC

Intrinsic parameters in camera calibration are commonly solved with high precision using the homography matrix of space coordinates and image coordinates of a template. However, in practice the accuracy in the extraction of corner points of the calibration plate is usually affected by the lighting environment, leading to obvious fluctuation in the calibration results. This study proposes a novel self-calibration method to improve the calibration accuracy and reduce the instability of calibration results by using circular points and the RANdom SAmple consensus (RANSAC) method. The circular points for the intrinsic parameters are calculated using the cross-ratio method with a calibration plate containing nine corner points. The distance between the circular points and image of the absolute conic is defined. The threshold value of the RANSAC model is simulated by a computer. The intrinsic parameters are initially estimated using the unreliable calibrating images excluded by the RANSAC method. The definition of the threshold is based on the Sampson estimation. The maximum likelihood estimation method is performed to reestimate the intrinsic parameters and optimize the calibration result. The findings of the numerical simulations and experiments on wing-fuselage docking based on monocular vision demonstrate that the proposed method is more robust and efficient at improving the calibration accuracy than the traditional methods. The measurement error is reduced to less than 0.013 mm when the calibration algorithm is applied to actual applications such as wing-fuselage docking.