面向大视场视觉测量的摄像机标定技术

Camera Calibration Technique of Wide-Area Vision Measurement

提出了一种面向大视场高精度视觉测量的摄像机标定新方法,该方法采用亮度自适应的单个红外发光二极管(IR-LED)作为目标靶点,将该靶点固定在三坐标测量机的测头上,并依次精确移动至预先设定的空间位置,每次靶点到达设定的空间位置时,摄像机对靶点进行图像采集。利用三坐标测量机的精确位移,在三维空间构成一个虚拟立体靶标。针对虚拟立体靶标在大视场摄像机标定中只能覆盖一小部分标定空间的问题,通过自由移动摄像机在多个方位对虚拟立体靶标进行拍摄,使得多个虚拟立体靶标分布于整个标定空间。摄像机在每个方位对虚拟立体靶标的拍摄都标定出一组摄像机的内、外参数,然后以摄像机内参数和摄像机在各个方位下拍摄的虚拟立体靶标在摄像机坐标系下的位置及姿态参数为优化变量,建立以所有三维靶点位置重投影误差平方和为最小的目标函数,采用非线性优化方法求解摄像机标定参数的最优解。该方法较好地解决了大视场视觉测量中大尺寸靶标加工困难、摄像机标定精度难以保证的问题。仿真和实际标定实验均证明此方法可以有效提高大视场摄像机的标定精度。

A new method is proposed for wide-area camera calibration using virtual calibration target created by one infrared Iight-emittng diode （IR-LED） marker fixed on the moving probe of a coordinate measuring machine （CMM）. The camera takes an image when the IR-LED marker moves accurately with the CMM probe to each preset position in turn. Since the measurement space of common CMM is not large enough, the virtual calibration target is imaged in several camera poses to cover the whole measurement volume of the wild-area camera. The calibration parameters are initially obtained by using the 3D information of the virtual target and its image taken from each camera pose. The calibration parameters are finally solved by nonlinear optimization of the intrinsic parameters and the positions and orientations of the virtual calibration target with respect to the camera. The proposed method requires no large-size calibration target and can ensure the calibration performance of the proposed method.