Huo Ju, Zhang Guiyang, Cui Jiashan, Yang Ming
Appl Opt. 2018 Jul 1;57(19):5492-5504. doi: 10.1364/AO.57.005492.
This paper is concerned with the precise calibration algorithm and error compensation technique for a binocular vision measurement system. By invoking a fixed constraint relationship between the binocular cameras, these two cameras can be preferably bound together as one camera, which means that in each iteration only one camera needs external parameter tuning for optimized encapsulation, and the iteration time is reduced. The global optimal solutions are obtained by substituting the initial iterative solutions into the optimized iterative equations of the spatial points. The derived improved iterative algorithm obtains the camera parameters with high precision and performs better in noise resistance. Moreover, the error compensation matrix is constructed to correct the relative pose of binocular cameras, upon which solutions to the error compensation matrix are solved via the unit quaternion. Consequently, the relative pose matrix is revised, allowing for further improvement in the calibration performance. Finally, simulation and practical experiments are demonstrated. The relevant results confirm the precision and robustness of the proposed approach.
本文关注双目视觉测量系统的精确标定算法和误差补偿技术。通过调用双目相机之间的固定约束关系，这两个相机可以被较好地绑定为一个相机，这意味着在每次迭代中只需要对一个相机进行外部参数调整以实现优化封装，从而减少了迭代时间。通过将初始迭代解代入空间点的优化迭代方程中获得全局最优解。推导得到的改进迭代算法能够高精度地获取相机参数，并且在抗噪声方面表现更好。此外，构建误差补偿矩阵来校正双目相机的相对位姿，通过单位四元数求解误差补偿矩阵的解。因此，相对位姿矩阵得到修正，从而进一步提高了标定性能。最后进行了仿真和实际实验。相关结果证实了所提方法的精度和鲁棒性。
