中文摘要：

现有线结构光多视觉传感器现场校准通常需要先完成线结构光视觉传感器局部标定,再进行多视觉传感器全局校准,效率低、且在搬运过程容易对线结构光视觉传感器局部标定结果造成影响,降低测量系统精度。针对以上问题,提出一种可以同步实现线结构光视觉传感器局部标定和多视觉传感器全局校准的现场同步校准方法。该方法以自由移动三次以上的一维靶标为中介,将两个无共同视场的线结构光视觉传感器联系在一起。以一维靶标特征点距离已知为约束条件,由交比不变性,得到所有一维靶标点及光条与一维靶标相交点在各线结构光视觉传感器图像坐标系下的图像坐标,进而求解出两个线结构光视觉传感器之间的转换矩阵和一维靶标特征点及光条与一维靶标相交点在全局坐标系下三维坐标,通过拟合求解全局坐标下的光平面方程,采用非线性优化算法求解出两个线结构光视觉传感器间的转换矩阵与光平面方程的最优解。如果视觉传感器多于两个,可通过两两校准方式实现多个视觉传感器现场校准。试验结果表明,该方法同步校准方均根误差可优于0.14 mm。

英文摘要：

Existing methods for calibrating multiple line structured light vision sensors often requires each structured light vision sensor to be locally calibrated before the global calibration of all vision sensors can be carried out. It is not only time consuming, but may also reduce the accuracy of the measurement system due to the structural change of individual vision sensors during the handling or assembling process after individual calibration. A method that can achieve simultaneously the local calibration of individual line structured light vision sensor and the global calibration of all vision sensors is presented. The two non-overlapping vision sensors are linked together through the 1D target, which is positioned in front of the two vision sensors for at least three times. Using the known distances between feature points on the 1D target and the cross-ratio invariance property, the image coordinates of the 1D target feature points and the intersection points of the 1D target and the structured light planes in the image coordinate frame of each vision sensor are computed, and the coordinate transformation between the two vision sensors is solved. Using the coordinate transformation, the spatial coordinates of the target feature points and the intersection points are computed in a global coordinate frame and the plane equation of structured light planes are computed by fitting the intersection points. The light plane equations and the coordinate transformation between the two vision sensors are optimized simultaneously by non-linear optimization. If there are more than two vision sensors, the calibration of multiple vision sensors may be achieved by the pair-ware. Experimental result shows that the RMS error of the proposed calibration is less than 0.14 mm.