一种基于线性模型的机器人手眼标定新方法

针对基于传统手眼标定模型AX=XB的标定问题,引入了由摄像机投影矩阵M构成的数学模型MY=M'YB,并针对该模型提出了一种新解法。该方法运用矩阵直积和特征向量理论,将关于机器人手眼关系矩阵X的问题描述转换成关于关联矩阵Y的线性方程,运用最小二乘法求得线性闭解。该方法消除了传统方法多次分解摄像机传感器系统的内外参数而引入的摄像机传感器系统的固有误差和随机误差的影响。实验表明该方法可以提高手眼系统标定精度。

For the problem of hand--eye calibration based on the classic transform equation of the form AX = XB, the paper presents a new approach to robot hand--eye calibration based on the projection matrix M that yields a homogeneous transform equation of the form My = MYB. This approach transforms the problem of hand--eye rnatrix X into linear problems of relating matrix Y by integrating matrix direct--multiply with matrix singular value method and solves the hand--eye matrix by lease --squares procedure. This method avoids the inherent errors and random errors of camera sensor system resulted from decomposition of the projection matrix M. Experimental results indicate that the proposed approach can improve the precision of hand--eye calibration system.