Research On Fast Camera Calibration Algorithm

The application of calibrated camera has been widely accepted. It is easy to implement vision measuring, 3D reconstruction based on a calibrated camera. But the current calibration method is still relatively complex, which impeded the further usage of the calibrated camera. It is absolutely needed for a new calibration method with accuracy and speed guaranteed, which is very easy to use.After researching of the developments of camera calibration, it is clearly that the key point of improving calibration algorithm is describing the camera imaging theory with a more general and accurate model as well as elevate the feasibility of calibration via degrading the requirements of the calibration auxiliary stuff. This work mainly focused on upgrading the feasibility of calibration as well as improving the calibration accuracy.Firstly, constrains out of the calibration points, a more e?cient converging is reached by this extra constrain. Because of the robustness of this constrain, the requirements of calibration points is reduced, that is less calibration board image is required.And the elevating of accuracy is certi?ed by the experiments.Secondly, models of camera lens distortion are investigated. Then, a new nonlinear template board based calibration method, based on the projection invariant of cross ratio of intersected lines, of lens distortion is proposed, which is able to be applied before the calibration of camera.Thirdly, theories of absolute conic in camera calibration is researched. Research shows that an absolute conic constrain can be acquired directly through the ordinary calibration boards. And based on it, accompanied with nonlinear optimization of reprojecting error, a full camera calibration can be achieved. Experiments certi?es the feasibility and accuracy of the proposed method.Finally, in order to evaluate the proposed calibration methods and related improvements, a new high-accuracy 3D measuring system based on vision measuring is established. The accuracy differences can be evaluated by checking the vision measuring results based on different calibration algorithm and comparing with the ground truth.Experiments shows that the improved calibration algorithm based on the height constrain owns the highest measuring accuracy.