基于3D视觉点云配准的高精度手眼标定方法

High precision hand-eye calibration method based on 3D visual point cloud registration

针对二维视觉测量方式缺乏Z轴深度信息导致手眼标定结果精度不高的问题,提出一种基于3D视觉点云配准的高精度手眼标定方法,搭建了一套单目结构光手眼标定系统,以传统的手眼标定模型为基础,采用最小化点云配准误差算法解算手眼标定矩阵。采用基于矩阵直积参数化方法初步求解手眼标定矩阵并作为初值,构建点云配准误差优化模型,以最小化点云配准误差为代价函数,同时,为保证手眼矩阵中旋转矩阵的正交特性,把代价函数从李群空间变换到李代数空间下,并对李代数空间下的代价函数进行最小二乘法迭代求解,获得最优的手眼标定矩阵。实验结果表明,所提方法具有较好的稳定性和精确性,手眼标定结果满足基于机械臂3D视觉测量的点云配准精度要求。

Aiming at the problem of low accuracy of hand-eye calibration results due to the lack of Z-axis depth information in two-dimensional visual measurement,a high-precision hand-eye calibration method based on 3D visual point cloud registration was proposed,and a set of monocular structured optical hand-eye calibration system was built.Based on the traditional hand-eye calibration model,the hand-eye calibration matrix was calculated by minimizing point cloud registration error algorithm.Using the matrix direct product parameterization method to initially solve the hand-eye calibration matrix and take it as the initial value,the point cloud registration error optimization model is constructed to minimize the point cloud registration error as the cost function.Meanwhile,in order to ensure the orthogonal characteristics of the rotation matrix in the hand-eye matrix,the cost function is transformed from Lie group space to Lie algebraic space.The optimal hand-eye calibration matrix is obtained by least squares iterative solution of the cost function in Lie algebraic space.The experimental results show that the proposed method has good stability and accuracy,and the hand-eye calibration results meet the requirement of point cloud registration accuracy based on the 3D vision measurement of the robot arm。