【摘要】 为提高双目相机标定的精度和效率，本文提出了一种基于新型编码立体靶标并结合高精度参数优化的双目相机标定方法。新型编码立体靶标通过整合4个空间姿态不同的编码平面靶标，只需拍摄一次即可完成双目相机标定，有效提高了标定效率；每个编码平面靶标内设置多个编码单元，在采集到局部靶标图像时仍能获得高精度标定结果；采用高精度参数优化方法，建立包含重投影约束、标准长度约束和共面约束的目标函数，有效提高了双目相机的标定精度。实验结果表明：相比于张氏标定方法，本文所提标定方法获得的左右相机平均绝对重投影误差分别降低了55.42%和57.22%，平均绝对标准长度误差降低了41.28%，平均绝对共面误差降低了63.04%。通过多次测量不同规格的标准量块，进一步验证了本文所提标定方法的可行性和有效性。更多还原

【Abstract】 Objective Binocular camera calibration is important to realize high-precision 3D measurements, dynamic target trajectory estimations, and 3D positioning through binocular vision. Zhang’s traditional calibration method based on a checkerboard target is currently the most widely used camera calibration method. However, Zhang’s calibration method has certain shortcomings. First,Zhang’s calibration method requires multiple shots of the checkerboard target, rendering the calibration process cumbersome. Second,when using the checkerboard target for calibration, the calibration accuracy is often degraded owing to the acquisition of local target images; this may sometimes also lead to failure of calibration. Finally, Zhang’s calibration method optimizes the parameters of a camera through reprojection constraints; moreover, the constraints are extremely singular, and hence, accurate camera calibration results cannot be obtained. To address the above-mentioned issues, this paper proposes a high-precision binocular camera calibration method based on a novel stereoscopic target.Methods To solve the problems encountered in Zhang stereoscopic target composed of four encoded plane targets with different spatial attitudes so that only one target image was needed to collect to complete the binocular camera calibration. In each coding plane target, multiple independent coding units were arranged to eliminate the traditional checkerboard structure, which could effectively address occlusion of the target. Second, each calibration corner was coded by a new variable-capacity coding flag, which improved the matching efficiency and accuracy of points with the same name. Finally, to enrich the camera parameter optimization constraints, a high-precision parameter optimization method was adopted, exploiting the three-dimensional information of the coding stereoscopic target and introducing standard length and coplanar constraints, which effectively improved the calibration accuracy of binocular cameras.Results and Discussions Compared with Zhang’s traditional calibration method, the method proposed in this paper demonstrates significantly improved calibration efficiency and accuracy. In Zhang’s calibration method, the checkerboard target must be shot multiple times, and each shot must ensure the integrity of the checkerboard target and the difference in the spatial posture of the checkerboard target. In contrast, the method proposed in this paper only requires to shoot the coding stereoscopic target once and can cope with the situation of the local target; therefore, the calibration efficiency is significantly improved. For the calibration accuracy,compared with Zhang’s calibration method, the proposed method reduces the reprojection errors of the left and right cameras by 55. 42% and 57. 22%, respectively; the standard-length error by 41. 28%; and the coplanarity error by 63. 04%(Fig. 6).Simultaneously, when using the standard gauge block for verification, the measurement error of the proposed method is reduced by 54. 61% compared with that of Zhang’s calibration method(Fig. 8). When the locally encoded stereoscopic target is captured, the relative error between the left and right camera reprojection errors is 7. 49%, the relative error of the standard-length error is 7. 42%,and the relative error of the coplanar error is 6. 36% compared with the complete target(Fig. 7). When a locally encoded stereoscopic target is recorded, the proposed method still demonstrates a high calibration accuracy for the binocular camera.Conclusions This paper proposes a binocular camera calibration method based on a coding stereoscopic target, combined with highprecision parameter optimization. The coding stereoscopic target includes four encoded plane targets with different spatial attitudes.By arranging multiple encoding units in the encoded plane target and encoding each calibration corner, the binocular camera can be calibrated based on a single target image acquisition. Compared with Zhang’s calibration method, which requires multiple shots of the checkerboard targets, the method proposed in this paper effectively improves the camera calibration efficiency. In this study, a highprecision parameter optimization method is used to establish an objective function that combines reprojection constraints, standard length, and coplanar constraints, thereby effectively improving the calibration accuracy of binocular cameras. The experimental results reveal that compared with Zhang’s calibration method, the proposed method effectively reduces the mean absolute reprojection error, mean absolute standard-length error, and mean absolute coplanarity error of binocular camera calibration. In addition, when the locally coded stereoscopic target is captured, the proposed method can still efficiently complete binocular camera calibration, and the relative error of each mean absolute error is less than 8% compared with that when the completely coded stereoscopic target is captured. This condition satisfies the requirements of high-precision calibration of binocular cameras for optical measurements.更多还原