针对手眼标定过程要求机器人运动次数过多的缺点,提出了一种改进的单特征点手眼自标定方法.引入手眼关系矩阵的解耦运算,分别标定手眼旋转矩阵和平移向量.运算过程无需计算特征点位置,操作过程仅需机器人末端有5次以上平移运动和2次以上旋转运动.实验与误差分析结果表明,所提方法满足工业机器人手眼视觉测量的需求. In order to overcome the shortcomings of the hand-eye calibration process requiring too many movements of the robot, an improved single-feature-point hand-eye calibration method is proposed. The hand-eye-relation matrix decoupling operation is introduced to calibrate the hand-eye rotation matrix and the translation vector respectively. At the point position, the operation only needs more than 5 translational movements and more than 2 rotations at the end of the robot.Experimental and error analysis results show that the proposed method meets the needs of industrial robotic hand-eye visual measurement.