基于主动视觉的机器人末端姿态测量  

Robot End-Effector Pose Measurement with Active Vision

作　　者：郭剑鹰[1] 吕恬生 

机构地区：[1]上海交通大学机械与动力工程学院,上海200030

出　　处：《上海交通大学学报》2003年第5期715-719,共5页Journal of Shanghai Jiaotong University

摘　　要：采用平面校准形及移动摄像机到不同形位摄取图像的方法来进行摄像机校准 ,由灭点正交性和单对应矩阵的固有特征估计摄像机内、外参数初值 ,通过非线性优化来最小化投影误差 ;利用遗传算法求解机器人手眼矩阵 ,避免了初值估计 ,可容易地得到机器人执行器末端相对于绝对坐标系的姿态 .实验和计算验证表明 ,这种测量方法是有效的 ,可以用于机器人校准、机器人辅助测量以及机器人辅助手术等领域 .A method for camera calibration using coplanar pattern was given. Many images are taken at different robot configurations by moving the robot-camera system freely. The initial values of intrinsic and extrinsic camera parameters are estimated through the orthogonality of the vanishing points and the intrinsic characteristic of the homographies. Then the nonlinear optimization method is implemented by minimizing the re-projection errors. The robot hand-eye transformation matrix is calculated through GA optimization. There is no need for the estimation of initial parameters. After camera calibration and robot hand-eye calibration, it is easy to obtain the robot manipulator end-effector pose relative to the world coordinate system (the measurement coordinate system). The proposed methods are verified through experiments and real data computation, which can be used for robot calibration, robot-aided measurement and robot-aided surgery etc.

关 键 词：机器人姿态测量 摄像机校准 手眼校准 平面形 单对应矩阵 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]