Parameter calibration of a vision sensor with a geometric similarity constraint

The precise parameters of vision sensors determine how accurately the device obtains its results. A novel effective approach, based on geometry similar constraint, has been derived. The proposed method decouples the lens distortion coefficients and linear parameters of the camera successfully. It is also able to estimate the camera's parameters with a small number of collinear points. All the distortion coefficients are solved through a geometry similar constraint model. Then corrected images are used to estimate linear parameters once distortion coefficients are acquired. Both a computer simulation and real-life experiment are carried out to validate the performance of the proposed method. The number of collinear points is first discussed with training data, and results showed the proposed method works better compared to currently widely used methods. The proposed method is also compared with current planar target based methods, which are commonly adopted in practical experiments. The results indicate that the proposed method is more reliable and effective.