为了实现双目立体视觉系统大范围高精度三维测量, 提出了一种大视场双目立体视觉系统柔性标定方法, 该方法将系统中各相机内部参数标定与相机间的姿态标定进行分离, 标定内部参数时, 只需要令标靶相对于相机任意摆放至少三个姿态, 对标靶上的编码标志点进行识别, 根据标靶上编码标志点信息, 建立各姿态下视图的对应关系, 粗略计算标志点的初始三维坐标; 建立多姿态下逆向投影误差最小的目标函数, 采用非线性最小二乘优化获取精确的相机内部参数和标志点三维坐标; 最后, 建立基于双相机逆向投影误差最小的目标函数, 优化得到精确的相机间姿态的外部参数。实验结果表明: 当测量空间为1 200 mm×1 000 mm×1 000 mm时, 立体视觉系统的测量精度优于0.1 mm, 满足大范围双目立体视觉系统的高精度测量需求。

In order to achieve high-precision three-dimensional measurement for binocular stereo vision system of large scale, a flexible calibration method in large field of view was proposed, which divided the calibration method into calibrating internal parameters and external parameters. To calibrate the internal parameters, the target was required to pose at least three posture corresponding to cameras, coded target on the target should be identified, according to the information of coded mark points on the target, a corresponding relationship among bottom views under all postures was established, initial three-dimensional coordinate of code targets were roughly calculated. The objective function of minimum reverse projection error under multi-pose was established, nonlinear least squares was used to obtain accurate intrinsic parameter of camera and three-dimensional coordinate of coded mark target. Finally, based on dual-camera, the objective function of minimizing reverse projection error was established, and the accurate extrinsic parameters of pose among cameras were optimized. The experimental result shows: the measurement space is 1 200 mm×1 000 mm×1 000 mm, measurement accuracy of the binocular stereo vision system is superior to 0.1 mm, which meet the requirements of high-precision measurement for binocular stereo visual system in wide vision field.