Objective Camera calibration is a key step to realize high precision three-dimensional(3D) visual measurement. In order to solve the problem of severe lens distortion effect and ensure the accuracy of camera calibration, lots of calibration research using different calibration objects is carried out. When using a one-dimensional reference ruler, due to the limited feature information of the ruler and insufficient dimensional information, a large number of calibration images need to be collected, resulting in low overall calibration efficiency. When using a 2D calibration board, although it can provide sufficient feature information, manufacturing large-sized calibration boards is not only costly, but also technically difficult in some cases. When using a 3D calibration object, the layout of measurement field of view is relatively complicated, and different measurement field of view require 3D calibration objects made with different scales and specifications, resulting in low calibration flexibility and poor applicability. To solve these problems, a large field of view camera calibration method based on self-made concentric circle array calibration board is proposed.

Methods In this method, a calibration board matching the measured field of view is made by using the concentric circle array pattern printed by laser to accurately extract the image coordinates of the concentric circle center. Then, based on the distribution characteristics of lens distortion on the imaging plane, the strategy of sub-region acquisition of calibration images is adopted to reduce the influence of large distortion imaging region on the estimation accuracy of camera internal parameters. At the same time, considering the size error between concentric circles and flatness error on the calibration board, an initial value algorithm for estimating internal parameters of camera, external parameters of camera and 3D coordinates of the calibration board is proposed. Finally, the system parameters are optimized by minimizing the back projection error of the image coordinates of the center of the circle as the objective function to improve the calibration accuracy of the camera.

Results and Discussions In order to prove the superiority of the proposed method, the back-projection errors calculated by the proposed method and ZHANG’s method were compared. Experimental results show that the root mean square error obtained by ZHANG’s method were 0.764 pixel and 0.806 pixel in the X and Y directions, respectively, while those obtained by the proposed method were 0.177 pixel and 0.201 pixel, which were reduced by 76.8% and 75.1%, respectively. In order to prove the validity of the proposed method, the measurement accuracy of the system calibrated by the proposed method was tested in the measuring field of 2.2 m × 2 m. It can be seen that the minimum measurement error of the system was 0.151 mm, and the average error was 0.478 mm. The minimum relative error was 0.016% and the mean relative error was 0.052%.

Conclusions Aiming at the technical requirements of camera calibration with large field of view, high precision and high applicability, a large field of view camera calibration method based on self-made concentric circle calibration board is proposed. This method uses the concentric circle array pattern printed by laser to make a calibration board for camera calibration, and proposes a high-precision calculation method of concentric circle center coordinates, which is used as the basis for model parameter estimation. Considering the distribution characteristics of lens distortion, the strategy of regional acquisition of calibration images is adopted, and considering the dimensional errors between concentric circles and the flatness errors of the calibration board, an initial value algorithm for estimating internal and external parameters of camera and 3D coordinates of the calibration board is proposed. In addition, the whole system parameters are optimized to achieve high precision camera calibration. Experimental results show that compared with ZHANG’s method, the root mean square error obtained by the proposed method is reduced by 76.8% and 75.1% in X and Y direction, respectively. The minimum relative error of the system is 0.016%, and the mean relative error is 0.052%, which proves the correctness and superiority of the proposed method.